| Original Full Text | University of Windsor Scholarship at UWindsor Electronic Theses and Dissertations Theses, Dissertations, and Major Papers 9-25-2024 Analyses of Spirituality, Social Support, Community Cohesion, and Resilience of Parental-Intimate Partner Violence Exposed Individuals Using Structural Regression Models Jenna Rose Emma Parsons University of Windsor Follow this and additional works at: https://scholar.uwindsor.ca/etd Part of the Clinical Psychology Commons Recommended Citation Parsons, Jenna Rose Emma, "Analyses of Spirituality, Social Support, Community Cohesion, and Resilience of Parental-Intimate Partner Violence Exposed Individuals Using Structural Regression Models" (2024). Electronic Theses and Dissertations. 9557. https://scholar.uwindsor.ca/etd/9557 This online database contains the full-text of PhD dissertations and Masters’ theses of University of Windsor students from 1954 forward. These documents are made available for personal study and research purposes only, in accordance with the Canadian Copyright Act and the Creative Commons license—CC BY-NC-ND (Attribution, Non-Commercial, No Derivative Works). Under this license, works must always be attributed to the copyright holder (original author), cannot be used for any commercial purposes, and may not be altered. Any other use would require the permission of the copyright holder. Students may inquire about withdrawing their dissertation and/or thesis from this database. For additional inquiries, please contact the repository administrator via email (scholarship@uwindsor.ca) or by telephone at 519-253-3000ext. 3208. i Analyses of Spirituality, Social Support, Community Cohesion, and Resilience of Parental-Intimate Partner Violence Exposed Individuals Using Structural Regression Models By Jenna Parsons A Thesis Submitted to the Faculty of Graduate Studies through the Department of Psychology in Partial Fulfillment of the Requirements for the Degree of Master of Arts at the University of Windsor Windsor, Ontario, Canada 2024 © 2024 Jenna Parsons ii Analyses of Spirituality, Social Support, Community Cohesion, and Resilience of Parental-Intimate Partner Violence Exposed Individuals Using Structural Regression Models by Jenna Parsons APPROVED BY: _________________ K. Nikolova School of Social Work ______________________________________________ L. Rappaport Department of Psychology ______________________________________________ P. Fritz, Advisor Department of Psychology September 9, 2024 iii DECLARATION OF ORIGINALITY I hereby certify that I am the sole author of this thesis and that no part of this thesis has been published or submitted for publication. I certify that, to the best of my knowledge, my thesis does not infringe upon anyone’s copyright nor violate any proprietary rights and that any ideas, techniques, quotations, or any other material from the work of other people included in my thesis, published or otherwise, are fully acknowledged in accordance with the standard referencing practices. Furthermore, to the extent that I have included copyrighted material that surpasses the bounds of fair dealing within the meaning of the Canada Copyright Act, I certify that I have obtained a written permission from the copyright owner(s) to include such material(s) in my thesis and have included copies of such copyright clearances to my appendix. I declare that this is a true copy of my thesis, including any final revisions, as approved by my thesis committee and the Graduate Studies office, and that this thesis has not been submitted for a higher degree to any other University or Institution. iv ABSTRACT Background: Intimate partner violence (IPV) refers to physical, emotional, and sexual harm and/or controlling behaviours by a current or former intimate partner. Children are often present when parental IPV occurs which puts them at a higher risk for many negative immediate and long-term physical, psychological, and social consequences. Despite childhood exposure to IPV, positive outcomes are possible which has led to the study of resilience. Resilience refers to the ability to cope with stress and thrive despite adverse experiences. Previous research has highlighted the importance of social supports (e.g., family, friends, and professional supports like IPV services) for resilience in adults exposed to IPV as children. Despite this, little is known about the roles that community factors and spirituality play in resilience in P-IPV exposed individuals. Objectives: To investigate the role of and relations among individual, interpersonal (i.e., social support, spirituality), community/organizational factors (i.e., community cohesion) using a social-ecological model of resilience for individuals exposed to IPV as children. Understanding the mechanisms that promote resilience in P-IPV exposed children can inform the creation of target driven interventions that may bolster resilience in adulthood for P-IPV exposed children. Methods: The current study examined data from respondents who reported exposure to parental intimate partner violence before the age of 15 years on the 2019 General Social Survey conducted by Statistics Canada. Results: As predicted, there were significant relationships between social support and resilience, and community cohesion and resilience, respectively. However, the overall model examining the relationships between social support, community cohesion, and resilience resulted in poor model fit. Significance: Findings from this study add to our current understanding of resilience in P-IPV exposed individuals by demonstrating that social support and community cohesion both significantly impact resilience. v DEDICATION This thesis is dedicated to the memory of Uncle Derrick Murray, whose love and light touched the lives of so many and has inspired me to carry on the torch. vi ACKNOWLEDGEMENTS Firstly, I would like to thank my supervisor Dr. Patti Timmons Fritz for her support and guidance over the past two years. Thank you for trusting in my abilities and supporting me throughout the many challenges that arose while completing my thesis. I appreciate your support in seeing my vision through and allowing me to choose a project that I am genuinely passionate about. Thank you for the many revisions you provided through the various drafts of my thesis document and for always quickly returning revisions so that we could meet tight deadlines. Next, I would like to thank my committee members, Dr. Rappaport and Dr. Nikolova for their thoughtful contributions and brainstorming during my proposal session and beyond. Your feedback and advice have been instrumental in strengthening my thesis. Another huge thank you to Dr. Rappaport for providing statistical expertise to help with my analyses. My methodology is much stronger due to your invaluable input. I would also like to acknowledge the Social Sciences and Humanities Research Council and the Health Research Centre for Violence Against Women for providing the funds to conduct this study. Thank you to Statistics Canada for providing me access to the 2019 GSS Victimization data. A huge thank you to Shane Goodwin from Statistics Canada’s Research Data Centre (RDC) for providing input and expertise on structural equation modeling. Also, to Mike Houlahan for always going the extra mile to support my data analyses at the RDC. Most importantly, I would like to thank my family and partner. This page is not long enough to express just how much your unwavering love and support mean to me. Thank you for supporting my victories and being a soft place to land when times are tough. I am honoured to be on the journey of becoming the first Dr. Parsons in our family. I love you beyond measure. vii TABLE OF CONTENTS DECLARATION OF ORIGINALITY .......................................................................................... iii ABSTRACT ................................................................................................................................... iv DEDICATION ................................................................................................................................ v ACKNOWLEDGEMENTS ........................................................................................................... vi LIST OF TABLES ......................................................................................................................... ix LIST OF FIGURES ........................................................................................................................ x LIST OF ABBREVIATIONS/SYMBOLS .................................................................................... xi CHAPTER 1 ................................................................................................................................... 1 INTRODUCTION .......................................................................................................................... 1 Impacts of P-IPV Exposure on Children ..................................................................................... 3 Psychological/Emotional/Behavioural Impacts of Childhood IPV Exposure ........................ 4 Social Impacts of Childhood IPV Exposure ............................................................................ 4 Cognitive/Intellectual/Academic Impacts of Childhood IPV Exposure .................................. 5 Physiological and Physical Impacts of Childhood IPV Exposure .......................................... 5 Moderators in P-IPV Exposure ................................................................................................... 6 Resilience .................................................................................................................................... 8 Resilience Science ................................................................................................................. 10 The Distinction Between Resilience and Post Traumatic Growth ........................................ 12 A Social-Ecological Approach to Resilience ........................................................................ 13 Resilience in Childhood Exposure to Intimate Partner Violence (IPV) ................................ 17 Protective Factors That Promote Resilience Among IPV Exposed Children ....................... 17 Community Factors and Social Support ................................................................................... 23 Spirituality and Social Support ................................................................................................. 23 Study Objectives ........................................................................................................................ 24 Research Questions (RQs). ................................................................................................... 24 Hypotheses. ........................................................................................................................... 25 CHAPTER II: METHODOLOGY ............................................................................................... 25 Participants ............................................................................................................................... 25 Design ......................................................................................................................................... 2 Measures ..................................................................................................................................... 2 Inclusion Criterion .................................................................................................................. 2 Predictor Variables ................................................................................................................. 3 Outcome Variable: The Conceptualization of Resilience for the Study .................................. 6 Data Analysis .............................................................................................................................. 6 Structural Equation Modelling ............................................................................................... 6 Original Model Specification .................................................................................................. 7 SEM Assumptions .................................................................................................................... 9 Correlations .......................................................................................................................... 10 Examining Model Fit ............................................................................................................. 11 CHAPTER 3 RESULTS ............................................................................................................... 12 viii Data Recoding and Cleaning .................................................................................................... 12 Assumption Testing ................................................................................................................... 16 Model 1R: Direct Relationship Between Social Support and Resilience .................................. 24 Model 1B Association Between Social Support and Resilience ............................................ 28 Model 1C Direct Effect Between Social Support and Resilience with Fixed Factor Loadings............................................................................................................................................... 32 Comparison of Models for Research Question 1: The Relationship between Social Support and Resilience ....................................................................................................................... 36 Question 2: Does Having a Higher Level of Spiritualty Correspond With Increased Resilience? ................................................................................................................................ 38 Model 2: A Structural Regression Model of Spirituality and Resilience .............................. 38 Research Question 3: Does Having a Higher Level of Community Cohesion Correspond With Increased Resilience? ............................................................................................................... 38 Model 3A: A Structural Regression Model of Community Cohesion and Resilience ........... 38 Model 3B Association Between Community Cohesion and Resilience ................................. 42 Model 3C Direct Effect Between Community Cohesion and Resilience with Fixed Factor Loadings ................................................................................................................................ 47 Comparison of Models for Research Question 3: The Relationship between Community Cohesion and Resilience ....................................................................................................... 51 Research Question 4: When Accounting for the Relationship Between Social Support and Community Cohesion, Does an Increase in Social Support and Community Cohesion Each Uniquely Correspond With Increased Resilience? ................................................................... 52 Model 4R: Social Support, Community Cohesion, and Resilience ....................................... 52 Model 4B Relationships between Social Support, Community Cohesion, and Resilience with Freely Estimated Factor Loadings ....................................................................................... 56 CHAPTER 4 GENERAL DISCUSSION AND CONCLUSIONS ............................................... 59 Model Specific Discussion ........................................................................................................ 60 Findings and Discussion for Research Question 1: Relationship Between Social Support and Resilience ....................................................................................................................... 60 Findings and Discussion Research Question 3: Relationship Between Community Cohesion and Resilience ....................................................................................................................... 63 Findings and Discussion Research Question 4: Relationships Among Social Support, Community Cohesion, and Resilience ................................................................................... 67 The Comparison of Relationships Between Social Support and Resilience and Community Cohesion and Resilience ........................................................................................................... 69 Overall Limitations ................................................................................................................... 70 Capturing P-IPV Exposure ................................................................................................... 70 Measuring Resilience ............................................................................................................ 70 Strengths .................................................................................................................................... 72 Practical Importance of Continued Research on Resilience in P-IPV Exposed Individuals .... 72 References .................................................................................................................................... 75 Appendices ................................................................................................................................... 95 Vita auctoris ............................................................................................................................... 103 ix LIST OF TABLES Table 1: Examples of Factors Operating at Each Level Within a Social-Ecological Model for Understanding Trauma and Its Effects (SAMSHA, 2014) ............................................................ 16 Table 2: Highest Education Level of Respondents ........................................................................ 26 Table 3: Ethnicities of Respondents .............................................................................................. 27 Table 4: Aboriginal Identities of Respondents ................................................................................ 1 Table 5: Relevant Variables from the 2019 General Social Survey ................................................ 4 Table 6: Shapiro Wilk Normality Test Values ................................................................................ 17 Table 7: Number of Physical P-IPV Exposures Reported by Respondents ................................... 19 Table 8: Frequencies of Responses for Survey Items included in Structural Regression Models 20 Table 9: Correlations Among Variables in Proposed Structural Models ...................................... 23 Table 10: Model 1R: Latent Variable Loadings and Variances .................................................... 27 Table 11: Model 1R: Squared Multiple Correlation (R2) Values for Each Exogenous Variable .. 28 Table 12: Model 1B: Latent Variable Loadings and Variances .................................................... 31 Table 13: Model 1B: Squared Multiple Correlation (R2) Values for Each Exogenous Variable .. 32 Table 14: Model 1C: Latent Variable Loadings and Variances .................................................... 35 Table 15: Model 1C: R2 Values for Each Variable ........................................................................ 36 Table 16: Comparison of Model Fit Statistics for Model Examining Research Question 1 ......... 37 Table 17: Information Criteria Values for Models Examining Research Question 1 ................... 37 Table 18: Model 3A: Latent Variable Loadings and Variances .................................................... 41 Table 19: Model 3A: R2 Values for Each Manifest Variable ......................................................... 42 Table 20: Model 3B: Latent Variable Loadings and Variances .................................................... 45 Table 21: Model 3B: R2 Values for Each Manifest Variable ......................................................... 46 Table 22: Model 3C: Latent Variable Loadings and Variances .................................................... 50 Table 23: Model 3C: R2 Values for Each Variable ........................................................................ 51 Table 24: Comparison of Model Fit Statistics for Model Examining Research Question 3 ......... 51 Table 25: Model 4R: Latent Variable Loadings and Variances .................................................... 55 Table 26: Model 4R: R2 Values for Each Manifest Variable ......................................................... 56 Table 27: Model 4 and 4b: Social Support, Community Cohesion and Resilience Eigenvalues .. 59 x LIST OF FIGURES Bar Graph Displaying the Percentage of Missing Data for Each Variable ................................. 14 Figure 2 ......................................................................................................................................... 16 Data Cleaning Procedures ............................................................................................................ 16 Figure 3 ......................................................................................................................................... 18 Q-Q Plot to Examine Univariate Normality ................................................................................. 18 Figure 4 ......................................................................................................................................... 25 Model 1 Revised: A Structural Regression Model of a Direct Relationship Between Social Support and Resilience ................................................................................................................. 25 Figure 5: Model 1B: A Structural Regression Model of the Correlational Relationship Between Social Support and Resilience ...................................................................................................... 29 Figure 6 ......................................................................................................................................... 33 Model 1C: A model of the Direct Relationship Between Social Support and Resilience with Fixed Factor Loadings ............................................................................................................................ 33 Figure 7 ......................................................................................................................................... 39 Model 3A: A Structural Regression Model of the Direct Relationship Between Community Cohesion and Resilience ............................................................................................................... 39 Figure 8 ......................................................................................................................................... 43 Model 3B: A Structural Regression Model of the Correlational Relationship Between Community Cohesion and Resilience ............................................................................................................... 43 Figure 9 ......................................................................................................................................... 48 Model 3C: A Structural Regression Model of Community Cohesion and Resilience with Fixed Factor Loadings ............................................................................................................................ 48 Figure 10 ....................................................................................................................................... 53 Model 4R: A Structural Regression Model of Community Cohesion, Resilience, and Social Support .......................................................................................................................................... 53 Figure 11 ....................................................................................................................................... 57 Model 4B: A Structural Regression Model of Community Cohesion, Resilience, and Social Support -Freely Estimated Factor Loadings ................................................................................ 57 xi LIST OF ABBREVIATIONS/SYMBOLS AIC Akaike Information Criterion CFI Bentler’s Comparative Fit Index IPV Intimate Partner Violence P-IPV Parental Intimate Partner Violence TLI Tucker Lewis Index RMSEA Root Mean Square Error of Approximation SEM Structural Equation Modelling SR Structural Regression SRMR Standardized Root Mean Square Residual 1 CHAPTER 1 INTRODUCTION Intimate partner violence (IPV) is a prevalent form of gender-based violence that includes sexual, physical, and psychological harm, and/or coercive control from a current or former intimate partner (Butchart et al., 2014). This could be a marital, common-law, dating, or casual partner (Scott et al., 2022). According to the World Health Organization (WHO), 27% of ever-married/partnered women worldwide aged 15 to 49 years have experienced lifetime of physical and/or sexual IPV (WHO, 2021). The Survey of Safety in Public and Private Spaces (SSPPS) conducted by Statistics Canada (2018) indicated that, in Canada, 44 percent of women who have been in an intimate partner relationship (roughly 6.2 million) reported experiencing sexual, physical, or psychological IPV since the age of 15. Of these 6.2 million Canadian women, 29 percent had experienced ten or more instances of sexual, physical, or psychological abuse (Cotter, 2021). By virtue of IPV estimates being largely based on population-based surveys, underreporting of IPV is probable; therefore, the global prevalence of IPV is possibly considerably larger than the previously cited estimates (Satyen et al., 2019). Due to the devastating immediate and long-term consequences, the WHO has declared IPV a major global public health concern (WHO, 2013). IPV impacts people of all genders; however, the majority of victim-survivors are women, and the violence is commonly perpetrated by men (Cotter, 2021). Of police reported IPV incidents in 2019, as in previous years, rates for women were 3.5 times higher than for men; 536 versus 149 per 100,000 population (Conroy, 2019). Because IPV disproportionately affects women, gendered language matching this dynamic will be used throughout. 2 Although parents often think that their children are unaware of violence in the home, research demonstrates that many children witness domestic violence assaults (Sinha, 2013). In 2018 in Canada alone, out of the 37,922 substantiated investigations of child maltreatment, 45% (17,051) were cases of childhood exposure to IPV (Fallon et al., 2015). Similarly, among Canadian women who were married or in common law partnerships and reported experiencing IPV within the past five years, 51% reported that they believed their children were exposed to the IPV (Statistics Canada, 2015). These statistics understate the widespread impact of IPV exposure due to the chronic underreporting of IPV (Public Health Agency of Canada, 2018). As such, in Canada, childhood exposure to IPV is recognized as a major public health concern. Parental IPV (P-IPV) refers to abusive or aggressive behaviours that occur between children’s parental figures and/or caregivers, regardless of their formal or legal parental status (Fritz & Roy, 2022). The majority of researchers consider exposure to P-IPV to occur when children are directly or indirectly involved in parental physical or sexual violence (Evans et al., 2008; Fritz & Roy, 2022). Children may be with their mother when she is demeaned or physically abused; hear loud conflict and violence; tend to their parent’s injuries after assault; be physically used as a weapon; tend to their own injuries and/or sequalae of trauma; call emergency services; manage interactions with a parent whose behaviour alternates between caring and violent; be abducted by the perpetrator; be present while their parent(s) is/are arrested; or are forced to leave home with a parent, leaving behind important social supports (i.e., school, peers, and family; Cunningham & Baker, 2007). Some previous literature has referred to children as witnesses of P-IPV. It is important to note, however, that when children live in a house with violence, they are never a passive recipient of these events. Although the children are never to blame for the violence, many actively assess their perceived roles in "causing a fight,” interpret and predict what is occurring/will occur, worry about the consequences, problem solve, 3 and/or take action to emotionally and physically protect themselves and their siblings (Cunningham & Baker, 2007). According to Cunningham and Baker (2007) there are various roles that children and youth play when exposed to parental IPV including: acting as a caregiver for their abused parent and younger siblings; being the abused parent’s confidant; being the perpetrator’s confidant and/or ally (the child may be forced to say demeaning things to the abused parent, etc.); acting as the perfect child (who excels in school, has perfect behaviour, and never seeks help for problems as a mechanism to avoid violence); acting as the referee during conflict to try and mediate and keep the peace; and finally, the child may be a scapegoat (i.e., their behaviour is to blame for the abuse). Thus, it is unsurprising that the impact of exposure to P-IPV on children often leads to negative sequalae that persists for the remainder of their lives. Impacts of P-IPV Exposure on Children The majority of research on the impacts of IPV on children has used a developmental psychopathology lens (Bogat et al., 2023). Using such a lens requires researchers to view later developmental trajectories that may be impeded by exposure to early life adversity (such as P-IPV exposure). In this framework, equifinality refers to the many developmental pathways that a person can have that ultimately all lead to the same positive or negative outcome. Whereas Multifinality refers to a host of factors or one particular factor, that can have heterogeneous effects on an individual’s development, which is largely context dependent (i.e., based on the environment; Bogat et al., 2023). The current study investigated the impact that various individual, interpersonal, and community factors have had on Canadian adults exposed to P-IPV during childhood with a particular focus on resilience. However, before summarizing the literature on resilience among individuals who have been exposed to P-IPV, the many negative impacts of P-IPV exposure are first reviewed. 4 Psychological/Emotional/Behavioural Impacts of Childhood IPV Exposure On average, children who have been exposed to P-IPV exhibit higher levels of internalizing (negativity focused inward including anxiety, over-controlled behaviours and inhibition) and externalizing (negative behaviours directed externally including aggression, antisocial, and under-controlled behaviours) behaviours compared to children without P-IPV exposure (Fritz & Roy, 2022). These externalizing and internalizing symptoms have been found in P-IPV exposed children with small to moderate effect sizes (Chan & Yeung, 2009; Fong et al., 2019; Fritz & Roy, 2022; Howell et al., 2016; Vu et al., 2016). Among P-IPV exposed children, these internalizing symptoms have been found to manifest into mental health problems, including dissociation (Kimball, 2016), posttraumatic stress (Bogat et al., 2023; Chan & Yeung, 2009; Ehrensaft et al., 2017; Galano et al., 2021), depression and anxiety (Bogat et al., 2023; Gardner et al., 2019; Kimball, 2016; Onyskiw, 2003), suicidal attempts or ideation (Onyskiw, 2003), and other psychological problems, such as attention and thought problems (Kitzmann et al., 2003). Externalizing behaviours found in samples of children exposed to P-IPV include disruptive behaviour, aggression, noncompliant and/or antisocial behavior (Howell et al., 2016; Kimball, 2016; Onyskiw, 2003), and risk-taking behaviour as teenagers (Bair-Merritt et al., 2006; McTavish et al., 2016). Social Impacts of Childhood IPV Exposure When compared to children who have not been exposed to P-IPV, exposed children often experience more interpersonal problems and demonstrate poorer social competence (Chan & Yeung, 2009; Gewirtz & Edleson, 2007). Relatedly, a meta-analysis by McIntosh and colleagues (2021) demonstrated that exposure to severe IPV had a negative impact on attachment security for one- to five-year-old children (r = -.23, p = .02). Importantly, this is not a prescribed negative consequence of IPV exposure, as increasing age and sensitive/positive parenting can contribute 5 to secure attachment, lessening the impact of IPV exposure on attachment (Bogat et al., 2023; Noonan & Pilkington, 2020). However, despite the promising effects of increasing age and positive parenting, problematic attachment has often been found to trigger a chain of negative outcomes, including later socioemotional problems such as issues with peer relationships (Bogat et al., 2023). Cognitive/Intellectual/Academic Impacts of Childhood IPV Exposure Exposure to P-IPV is related to cognitive delays, executive dysfunction, attention issues, decreased memory function, and lower IQ (Fritz & Roy, 2022; Howell et al., 2016; Onyskiw, 2003). One mechanism for these deficits is chronic stress, which disturbs the natural balance of children’s stress biology systems (i.e., due to the constant overstimulation of certain brain structures, children’s arousal is elevated which hinders their ability to react appropriately to stress) and portends to abnormal structural and functional neurological development (Perry, 1997; Margolin & Gordis, 2003). This cascade triggered by prolonged stress impacts attention, learning, and memory (Fritz & Roy, 2022; Margolin & Gordis, 2000). Further negative sequalae of P-IPV exposure includes lower academic achievement in reading, language, and mathematics; lower school engagement; frequent absences and truancy; grade retention; and more general learning issues (Fritz & Roy, 2022; Howell et al., 2016). Proposed mechanisms for the negative sequalae at school include missing school or being too anxious, depressed, or traumatized, which interferes with a child’s ability to learn the material presented, leading to subsequent poor academic functioning (Bogat et al., 2023; Fritz & Roy, 2022; Supol et al., 2021). Physiological and Physical Impacts of Childhood IPV Exposure As previously alluded to, children exposed to P-IPV may be directly physically impacted by the violence when trying to intervene in the altercation (e.g., receive injuries from being punched, hit, etc.) or when one parent inflicts harm upon the child as a way of controlling or 6 distressing the other parent (Fritz & Roy, 2022). In extreme cases, the child may be at risk of death (Jaffe et al., 2012; McTavish et al., 2016). Indirect physical health consequences include gene level impacts (e.g., telomere erosion, gene expression, epigenetics – such as changes to DNA), somatic complains (e.g., headaches, body aches), impeded physical growth, fatigue, and vision, speech, and hearing problems (Fritz & Roy, 2022; Onyskiw, 2003). Exposed children are also more likely to have gastrointestinal issues, allergies, and asthma (Howell, 2011; Onyskiw, 2003). Another mechanism through which IPV exposure may impact physical health is increased risk-raking behaviour in adolescence, such as risky sexual behaviour and substance use (Fritz & Roy, 2022). Moderators in P-IPV Exposure A one-size-fits-all approach cannot be used to understand the negative impacts of childhood P-IPV exposure that have been reviewed above. It has been found that there are moderators that determine which children are more likely to experience the negative sequalae of P-IPV exposure. Some of these moderators exist at the level of the individual, including the child’s age, which has mixed results suggesting that impacts are worse for younger children (Gewirtz & Edleson, 2007; Onyskiw, 2003), for older children (Onyskiw, 2003; Sternberg et al., 2006), or no difference (Chan & Yeung, 2009; Evans et al., 2008); and the child’s sex suggesting that girls demonstrate more internalizing problems whereas boys exhibit more externalizing problems (which also has mixed results making it difficult to conclude if there are differential impacts based on sex; Fritz & Roy, 2022). Appraisals of the violence is another moderator, including the degree to which exposed children believe they are to blame for the violence as well as their ability to cope (Grych & Fincham, 1990). There are mixed results for abuse status as a moderator given that many children who are exposed to P-IPV also experience child abuse, which is referred to as the double whammy, and often leads to worse outcomes for children 7 (Chan & Yeung, 2009; Fritz & Roy, 2022; Kitzmann et al., 2003). Finally, socioeconomic status also has had mixed results (Martinez-Torteya et al., 2009) with some researchers concluding that children from low-SES families (less than $20,000/year) may be at increased risk of exposure to P-IPV (Bogat et al., 2023; Ernst et al., 2007; Wood & Sommers, 2011). There appears to be an additional moderator on low SES, the level of perceived social support, with higher levels of this variable mediating the effects of conflict in the home and being associated with fewer emotional and behavioural problems (Owen et al., 2008). Additionally, there are moderators that exist in a larger context of the child’s relationships. For instance, one moderator includes exposure to multiple types of violence, referred to as polyvictimization, which is associated with worse outcomes than P-IPV exposure alone (Bogat et al., 2023; Chan et al., 2017). Greater severity or increased frequency of violence, chronic P-IPV exposure, and physical IPV are similarly correlated with worse outcomes (Fritz & Roy, 2022; Grych & Fincham, 1990; Howell, 2011; Kitzmann et al., 2003). Next, caregivers and their characteristics are moderators as positive attachment with a warm and caring parent or other caring adults (e.g., neighbour, grandparent, coach, older sibling) is an important protective factor for P-IPV exposed children (Fritz & Roy, 2022; Howell, 2011). Finally, the adversity package is a set of factors that children exposed to P-IPV often experience, which is comprised of parental mental health problems and/or substance use, homelessness, poverty, criminality, social isolation, a large family, unemployment, and child abuse (Fritz & Roy, 2022; Holt et al., 2008). Overall, the literature demonstrates that there are a multitude of factors and moderators with complex relationships that must be considered when analyzing the potential negative sequalae of childhood P-IPV exposure. Given this, analyses using statistical models like the structural regression models used in the current analyses are helpful tools to elucidate the connection 8 between IPV exposure and potential mediators (here spirituality, social support, and community factors) with resilience. Resilience Despite childhood exposure to IPV, positive outcomes are possible, which has led to the study of resilience in IPV (Masten, 2001). One of the major difficulties in conducting research on resilience is that the field lacks a universal definition and conceptualization of resilience. Resilience has previously been conceptualized as an outcome, process, or trait (Fletcher & Sarkar, 2013). Some operational definitions of resilience include evidence of strong mental health despite adverse experiences (Herrman et al., 2011). The definitional debate is important as the way in which resilience is defined, has important implications for how the research is conducted (i.e., the theoretical boundaries that determine the nature, direction, and exactness of the research; Fletcher & Sarkar, 2013). Resilience is understood as a concept that encompasses a breadth of experiences, but most often refers to the ability to demonstrate positive adaption despite having experienced adversity (Luthar et al., 2000). Interestingly, resilience itself is not the end goal but rather is instrumental to decrease vulnerability and promote long-term sustainability (Serfilippi & Ramnath, 2018). When considering the many definitions that exist, there are essential elements of resilience: first to demonstrate resilience, a person needs to have experienced “a significant threat to their development,” (Masten, 2001, p. 229). Although positive stress is an important factor for healthy development, people are most likely to acquire resilience when the adverse events they experience are accompanied by strong, frequent, or prolonged stress, or when the effects are buffered by supportive relationships (Herrman et al., 2011). Accordingly, it is impossible for a person to demonstrate resilience if they have never experienced severe adversity or had been exposed to a significant threat. Secondly, positive adaptation needs to be achieved despite having 9 experienced or currently still experiencing such adversity (Fletcher & Sarkar, 2013; Masten et al., 1990). Finally, the phenomenon of resilience is viewed as a process of recovery that occurs over time (Anderson & Bang, 2012; Ungar, 2013). In line with the common definitional elements of resilience, there have been two functional indicators of resilience used within the literature. The first is the assessment of adversity, challenge, or trauma such as a singular acute traumatic event or more chronic exposure to early adversity. The second is the assessment of subsequent psychological functioning among youth, such as behavioural outcomes, symptoms of depression or posttraumatic stress, and indicators of wellbeing (Stein et al., 2019). For the current study, resilience will be conceptualized using these functional indicators of resilience. The traumatic event under consideration will be the retrospective reporting of exposure to P-IPV before age 15. Second, a measure of participants’ subsequent psychological functioning and/or self-reported mental health will be used. Resilience has been defined in the context of P-IPV research as “a process of navigating through adversity, using internal and external resources (personal qualities, relationships, and environmental and contextual factors) to support healthy adaptation, recovery, and successful outcomes over the life course,” (Alaggia & Donohue, 2018, pp. 3–4). As such consistent with the most common conceptualization of resilience and the definition of resilience used within the P-IPV literature, resilience in this study will include factors related to positive psychological functioning as well as the absence of psychopathology examined using a socio-ecological framework. The socio-ecological framework seeks to understand the complexity of violence by examining the interactions and interrelations of many factors related to an individual and the contexts in which they operate. Four levels or contexts that contribute to violence are analyzed 10 using this model: individual, interpersonal, community/organizational, and societal (Baker et al., 2016). Theorists speculate that resilience is experienced due to individuals’ internal senses of being valued, feelings of security, and meaningful connections with people in their life (Kralik et al., 2006). The study of psychological resilience seeks to determine which elements are responsible for helping some people withstand or even thrive despite adversity, whereas other people do not (Fletcher & Sarkar, 2013). Thus, one aim of the current study is to examine the direct and mediating roles that meaningful connections and support (herein referred to as social support) play in the lives of individuals with childhood P-IPV exposure. Resilience Science More than half a century ago, resilience science was born. This subdiscipline grew out of a multidisciplinary curiosity of psychologists, psychiatrists, and pediatricians to understand the origin of the strikingly different outcomes of children who experienced disadvantage and adversity during childhood. According to Masten (2013), early researchers aimed to inform their practice by understanding the factors responsible for some individuals’ thriving whereas others floundered in the face of similar adverse experiences. A major contribution of this research field is the shift of frameworks from a deficit-focused orientation to models that are grounded in promotive and protective factors, positive aims, and adaptive capacities in multiple academic disciplines (Masten, 2011). Over time, researchers have identified an important kind of moderating effect, termed differential susceptibility (Belsky et al., 2007; Ellis & Boyce, 2011) or sensitivity to context (Boyce & Ellis, 2005) to refer to the impact of context in determining whether the same factors/characteristics serve as protective or vulnerability factors. Resilience research has experienced three previous phases – the first was focused on case studies; the second sought to investigate explanatory processes for positive outcomes; and the 11 third was focused on intervention research (Masten, 2015). In a sense, the field of resilience science has come full circle as the current fourth wave has been devoted to understanding the processes of resilience. However, the fourth wave brings along with it an integrative approach not previously used – this integration seeks to understand resilience as a complex, multi-level interconnection of systems from molecules to sociocultural factors (Masten, 2007). The fourth wave of research has been facilitated by huge advances made in computational and statistical approaches for data analyses (Masten & Cicchetti, 2016). Consistent with the fourth wave, the current study will use structural regression models to pool together data and determine how three sets of protective factors: spirituality, social support, and community cohesion impact resilience (i.e., positive outcomes) in a nationally representative sample of adults who were exposed to P-IPV during childhood. The initial waves of resilience science used two primary research strategies: person-focused and variable-focused strategies (Masten, 2001). Person-focused strategies included the comparison of individuals in high-risk groups – some of whom demonstrated successful adaptive function whereas others, did not; case studies; and investigation of pathways of adaptive function with the person at the center of the analysis. Variable-focused methods examine the relationships among the variables that are thought to be principle for resilience (including potential explanatory variables, variables that capture risk or adversity, and adaptive outcomes). These relationships may include direct or indirect effects (e.g., mediation or moderation) between risk and resilience (Masten & Barnes, 2018). Some studies combine the above-mentioned strategies which provide complementary strengths for the analyses. Person-oriented methods take a wholistic approach in viewing a person’s function as a system, whereas variable-oriented methods permit researchers to probe the specific underlying processes (Masten, 2001). Contemporary multivariate strategies (e.g., growth mixture modeling) are examples of 12 combining person- and variable-oriented approaches as people are grouped based similar trajectories of at least one variable, to identify broader patterns of adaptive functioning for the group at large (Muthén & Muthén, 2000). Given that the current study will be using structural regression modeling, resilience will be analyzed in a way that aims to identify broader patterns of adaptive functioning for a Canadian sample of adults who were exposed to P-IPV during childhood. The Distinction Between Resilience and Post Traumatic Growth Some children improve in functioning after experiencing a traumatic event such as being abused or exposed to P-IPV via a process referred to as Post Traumatic Growth (Tedeschi & Calhoun, 2004). For some individuals, this improvement has been found to be deeply profound (Tedeschi & Calhoun, 2004). Posttraumatic growth (PTG) has been defined in previous literature as the positive psychological change that a person experiences despite experiencing difficult life circumstances (Tedeschki & Calhoun, 2004). Previous research has suggested five domains of PTG including a new appreciation of life, personal strength (realization that one is stronger than they realized), re-examining one’s priorities, new possibilities (newly developed interests often connected to the trauma they experienced. E.g., leveraging lived experience as expertise by changing careers to working with individuals that had similar traumatic experiences), and a deeper spiritual connection and meaning in one’s life (Calhoun & Tedeschi, 2014). There are important conceptual differences between resilience and post-traumatic growth that must be highlighted. Recall that resilience is best understood as positive psychological functioning combined with the absence of psychopathology. Resilience differs from post-traumatic growth as resilience encapsulates positive psychological functioning despite exposure to adversity that is not an improvement of pre-exposure to trauma functioning. Post traumatic growth on the other hand, occurs when people who have experienced a traumatic event have 13 improved psychological functioning (in the form of the five domains mentioned above from Calhoun & Tedeschi, 2014) from pre-trauma exposure. So, in essence, resilience is comprised of positive psychological functioning in spite of adversity, whereas post traumatic growth is improved functioning as a result of the trauma. A Social-Ecological Approach to Resilience Some authors believe resilience is a personal strength that a person may harness to cope with adverse experiences (Black & Ford-Gilboe, 2004). However, most authors argue that the concept of resilience is not a static character trait; rather, it develops as a dynamic interaction of many factors in the context of experiencing stress (Lee et al., 2004). Understanding resilience solely at the individual level risks further pathologizing individuals who continue to struggle while experiencing adversity and creates a narrative that some individuals are unequipped to overcome adversity (Masten et al., 1990). Lacking resilience is not a character flaw, nor is it due to a lack of initiative on the part of the individual (Masten & Cicchetti, 2016). For these reasons, Masten cautions against using “resiliency” as it carries the built in connotation of a personality trait (Masten et al., 1990). Relatedly, an individualist perspective of resilience does not support the illumination of the mechanisms that underlie resilience, which can be used to guide appropriate interventions to bolster resilience (Masten et al., 1990). Starting at the biological level, individual adaptation that leads to resilience arises from many intersecting processes within the organism (e.g., at the hormonal, genetic, etc. level) as well as at the behavioural level (e.g., the capacity for learning and attention). More pronounced for resilience are the interactions between a person and the environments in which the person is embedded (influences from school, the community, or relationships with caregivers and peers; Jenkins, 2008). When referring to beneficial outcomes under adverse circumstances the term used is protective. However, when referring to negative outcomes under adverse circumstances, 14 it is a vulnerability factor (Jenkins, 2008). Protective and vulnerability factors that contribute to resilience operate at many intersecting levels of influence: individual, family, culture, and community (Herrman et al., 2011). Previous research has demonstrated that at the microenvironmental level, social supports, including strong relationships with friends and family, are positively linked with resilience (Herrman et al., 2011). Additionally, on the macrosystemic level, community factors (e.g., lack of exposure to violence, community services, sports and artistic opportunities, cultural factors) all have important contributions to resilience (Herrman et al., 2011). Resilience is best understood using a social-ecological framework (Anderson & Bang, 2012; Ungar, 2013) as resilience is viewed as a common capacity that occurs from the myriad of interactions between individuals and their respective contexts from the molecular level all the way to the macro-system level of ecology, society, and culture (Masten & Barnes, 2018). The social-ecological framework highlights the dynamic aspect of resilience that emerges from the interchange of individual, event, and environmental factors. Given the many intricate relationships that exist between three sets of protective factors (which serve as the main variables of interest for the current study: spirituality, social support, and community cohesion), a social-ecological framework has been loosely used to conceptualize the proposed structural regression models for this study. The specific socio-ecological framework that is used to organize the variables in the current study is the Substance Abuse and Mental Health Services Administration’s (SAMHSA) Social-ecological Model for Understanding Trauma and Its Effects which identifies individual, interpersonal, and community and organizational factors that serve as a model for both trauma and health using an intersectional and life course perspective. For examples of factors operating at each level within the socio-ecological model and its effects, see Table 1 (SAMSHA, 2014). 15 There are seven levels of SAMSHA’s social-ecological model. However, for the current analyses, only three levels of the model will be examined: individual, interpersonal, community and organizational. 16 Table 1 Examples of Factors Operating at Each Level Within a Social-Ecological Model for Understanding Trauma and Its Effects (SAMSHA, 2014) Individual Factors Interpersonal Factors Community and Organizational Factors Societal Factors Cultural and Developmental Factors Period of Time in History Age Family interactions Neighbourhood quality Laws Collective or individualistic cultural norms Societal attitudes related to family violence Biophysical state Peer interactions School system and/or work environment Provincial and Federal economic and social policies Ethnicity Changes in diagnostic understanding between editions of the Diagnostic and Statistical Manual of Mental Disorders Mental health status Significant other interactions Social services, family violence services, and health services quality and accessibility Media Cultural subsystem norms Personality traits (e.g., temperament) Parent/family mental health Faith-based settings Societal norms Cognitive and maturational development Education Parents’ history of trauma Transportation availability Judicial system Gender Social network Community socioeconomic status Coping styles Community employment rates Socioeconomic status Note. Table adapted from Baker, L., Straatman, A.L., Etherington, N., O’Neil, B., Heron, C., & Sapardanis, K. (2016). Towards a conceptual framework: Trauma, Family Violence and Health. London, ON: Knowledge Hub, Learning Network, Centre for Research & Education on Violence against Women & Children. https://www.learningtoendabuse.ca/resources-events/pdfs/Framework_paper-April-20171.pdf Mapping the variables of the current study onto SAMSHA’s social-ecological framework allows the results to be interpreted and compared to this model. As each level of the model offers a key point of intervention, the results of the current study can be interpreted within these levels 17 to determine points of intervention for IPV exposed children to help foster resilience. This allows for easier knowledge translation of results which will enable adults who interact with children (e.g., parents, teachers, social workers, relatives, psychologists) to intervene in the lives of children who have been exposed to IPV during childhood in ways that will foster resilience. Resilience in Childhood Exposure to Intimate Partner Violence (IPV) As previously mentioned, although parents often think that their children are unaware of violence in the home, research demonstrates that many children witness intimate partner violence assaults (Sinha, 2013). Childhood exposure to IPV encompasses direct exposure to violence (i.e., the child witnesses or experience the violence as it happens) and indirect exposure to the violence which is sometimes referred to as inferred exposure (Latzman et al., 2017). Examples of inferred exposure include hearing the violence happen from another room, seeing injuries on the caregiver that resulted from the violence, or knowing that their father is in jail due to assault charges (Tabibi et al., 2020). Of note, childhood exposure to IPV is included in the definition of child maltreatment in some jurisdictions (Black et al., 2020). A literature review from 2013 concluded that 26 to 50 percent of children who had been exposed to IPV had comparable outcomes as control participants (Laing et al., 2013). The question of why some children do well despite IPV exposure whereas others do not, has sparked resilience research in IPV exposed children. The overarching goal of such an investigation is to gather information that can be used in practice to increase the efficacy of prevention and intervention efforts for this population (Tabibi et al., 2020). Protective Factors That Promote Resilience Among IPV Exposed Children Previous research has identified many protective factors that may promote resilience in IPV exposed children. The term protective factor is used to refer to factors that are correlated with positive adaptation or outcomes in high adversity or high-risk situations (Masten, 2001). It 18 is important to note that these factors exist on many levels of the social-ecological framework (i.e., individual-, interpersonal-, and community and organizational-levels) and how a child responds to these protective factors is largely dependent on the complex interplay of factors and their interaction with children’s experiences of relational and structural violence or adversity. Because of the range of factors and their interactional nature, it is suggested that resilience be conceptualized and assessed using a multidimensional approach (Masten, 2001). Relatedly, promoting these protective factors may serve as important support for children’s health and well-being and to help facilitate resilience. Children benefit most when research, services, and supports operate using a socioecological framework that considers various strengths, availability of potential resilience sources, and promotive factors of each individual child (Tabibi et al., 2020). It was proposed that the current study would focus on three sets of protective factors: social support, spirituality, and community cohesion, that were to be included in the structural regression models. Social support has the strongest evidence as a mediator for resilience in P-IPV exposed children. There have been mixed results within the very limited research on the impacts of spirituality as a mediator of resilience and P-IPV exposure (Anderson & Bang, 2012; Howell & Miller-Graff, 2014; Tabibi et al., 2020; Yick, 2008; Yule et al., 2019). Similarly, there is little evidence to support community cohesion as a mediator in the indirect relationship between P-IPV exposure and resilience. Thus, the proposed models were created to analyze the direct relationships between social support and resilience; spirituality and resilience; community cohesion and resilience; and the correlations between the predictor variables (i.e., social support, spirituality, and community cohesion) for P-IPV exposed individuals. Individual Factors. Spirituality is an individual factor. Yule and colleagues’ (2019) conducted a meta-analysis of studies that included samples of children and adolescents who were 19 exposed to violence to examine the strength of the associations between protective factors and positive adaptation for this population. Yule et al. (2019) demonstrated promising protective effects of religious involvement (which was measured as whether a person was involved in a religious institution or not, as well as their religious beliefs and practices). Pooled effect sizes indicated that religious involvement was small but significantly related to resilience when examining cross-sectional (r = .05, p < .001) and longitudinal studies (r = .16, p < .05). The benefits of religious organization involvement have been attributed to the built-in supportive network of people with shared beliefs and values, which cultivates spiritual growth, and in turn, improves health and functioning in adults (Howell & Miller-Graff, 2014; Tabibi et al., 2020). In addition, Anderson and colleagues (2012) conducted a mixed-methods study to examine the recovery of 37 women who were formerly in abusive intimate partner relationships. In the qualitative interview, 31 women (83.8% of the sample) discussed that connection to a Higher Power (described by 24 women (64.9% of the total sample) as organized and informal practice of religion) after IPV exposure fostered a sense of value in life, meaning, and purpose, which helped them recover from the trauma they experienced. Further, roughly half of the women reported that the church played a pivotal role in their experience of trauma post-IPV exposure by providing them with practical assistance (e.g., housing and financial help), a sense of belonging, security, and emotional comfort (Anderson & Bang, 2012). The majority of survivors endorsed deriving strength from their personal spiritual connection (i.e., spirituality) rather than organized or institutional religious practices (i.e., religion; Yick, 2008). Research suggests that spiritual beliefs may serve to create meaning, comfort, and hope (Koenig, 2010). A review of six qualitative studies on IPV found that survivors of IPV used their spiritual beliefs as a source of strength to survive and cope with the IPV (Yick, 2008). Barring the aforementioned Yule and colleagues (2019) study, there appears to 20 be a dearth in the literature surrounding the connection between spirituality and resilience in individuals that were exposed to P-IPV during childhood. Based on the preponderance of evidence for the connection between spirituality and IPV outcomes for IPV survivors, it is hypothesized that spirituality also serves as a factor that promotes resilience for P-IPV exposed children. As such, the current study will focus on the impact of spirituality (which will be measured using the religion/spirituality of the person, frequency of formal spiritual practice, frequency of informal spiritual practice, and importance of religion) on resilience. In the current study, the direct path between spirituality to resilience will be analyzed to determine if spirituality is a predictor of resilience in P-IPV exposed children. Interpersonal Factors. Experiencing IPV is a huge life stressor that often comes with negative consequences for P-IPV exposed children, including frequent moves, social and financial strain, contact with the legal system or police, and separation from loved ones (Nikolova et al., 2021; Wathen & MacMillan, 2013). IPV may also have detrimental effects on a mother’s ability to parent their child(ren) as the trauma that they are experiencing/have experienced may lend itself to more neglectful and aggressive parenting (Kelleher et al., 2008). Mothers who experienced IPV have been found to be less emotionally warm toward their children and less involved in their children’s care and education (Margolin & Gordis, 2003). Mothers who have experienced IPV are at an increased risk of developing PTSD and depression (Levendosky et al., 2004). Further, maternal depression has been found to negatively impact a child’s functioning (English et al., 2003). In Yule and colleagues’ (2019) meta-analysis of protective factors for children who had been exposed to violence, they concluded that sources of social support were significant protective factors for resilience in IPV exposed children (i.e., the weighted effect sizes pooled from cross-sectional studies were r = .16, p < .001, for family support and r = .13, p < .01, for 21 peer support, and r = .18, p < .001, from longitudinal studies for family support). Similarly, a loving and supportive relationship with an adult (e.g., neighbour, teacher, coach, parent, grandparent) has consistently been found to be a vital protective factor to promote children’s resilience (Osofsky, 1999). These supportive adult-child relationships provide children with protection and scaffolding, which serve as essential building blocks of key capacities required to adaptively respond to adverse experiences (Osofsky, 1999; Tabibi et al., 2020). Moreover, a study by Coker and colleagues (2003) was conducted on 191 Black and White women survivors (aged 18 to 65) of physical IPV recruited from two university family practice clinics in the United States. Coker et al. (2003) used structural equation modeling to examine the impacts of social support (measured using the Social Support Questionnaire – Short Form; Sarason et al., 1987) on mental and physical health. The authors found that social support significantly mediated the relationship between physical IPV and better physical (B = -0.23, p < .01) and mental health (B = -0.27, p < .001), respectively. It is important to note, however, that this study was underpowered as it only included 191 participants and the growing consensus is that for SEM the samples should be of 200-400 for the most stable estimates across the various fit indices (Jackson, 2003). Additionally, Coker et al. (2003) focused on survivors of IPV, not P-IPV exposed children. Given the significant relationship found in the Coker and colleagues (2003) study using SEM (which was used in the foregoing study) and the previous evidence to demonstrate the promotive role of social support in resilience for children exposed to P-IPV, I hypothesize that there will be a significant positive relationship between social support and resilience in the proposed study for individuals exposed to P-IPV. For the proposed study, social support was a construct comprised of close relationships, community belonging, and family trust. 22 Community and Organizational Factors. At the community level, limited research has focused on potential protective factors. Community cohesion includes the presence of trustworthy, helpful, and involved neighbours (Herrero & Gracia, 2007). Research has demonstrated that community cohesion acts as a source of social support for children and their families (Sampson et al., 2002). Despite this, there is currently a dearth of research on the impact of community factors in IPV research (Yule et al., 2019). The limited research that has examined community cohesion as a protective factor for children who have experienced violence has been mixed. Yule and colleagues (2019) demonstrated that community cohesion is a significant protective factor for resilience in children exposed to violence (i.e., the weighted effect size calculated from three cross-sectional studies was significant, r = .13, p < .05). Yet, when examining data from four longitudinal studies, the weighted effect size was 0. Yule et al. (2019) proposed that this was due to the large variability among study effect sizes. In addition, Riina (2021) examined the relationships between neighbourhood cohesion and internalizing and externalizing behaviours in P-IPV exposed children. Using multilevel modeling, Riina (2021) found that there was a positive association between P-IPV and internalizing problems among youth in less cohesive neighbourhoods (γ = .13, SE = .06, p < .05). However, there was a nonsignificant relationship between internalizing problems for P-IPV exposed youth in more cohesive neighbourhoods (γ = −.05, SE = .07, p > .05). This suggests that higher levels of neighbourhood cohesion may act as a protective factor by mitigating the negative relationship between P-IPV exposure and internalizing problems. Riina (2021) also found that for youth from less cohesive neighbourhoods, there was a significant positive association between P-IPV exposure and externalizing behaviour (γ = .23, SE = .07, p < .01). However, for youth in more cohesive neighbourhoods, the association between P-IPV and externalizing problems was nonsignificant (γ = −.02, SE = .08, p > .05). Again, this suggests that 23 neighbourhood cohesion may buffer the negative effects of P-IPV exposure on externalizing problems. Given the mixed findings, the current study will analyze the relationships between community cohesion and resilience using a nationally representative sample to help shed light on the nature of the effect. Community Factors and Social Support It is hypothesized that the relationship between community cohesion (i.e., neighbour trust, neighbour helpfulness, neighbour connection) and resilience for individuals who were exposed to P-IPV as children will be mediated by social support. Among adolescents exposed to violence, community cohesion has been found to bolster resilience (Fagan et al., 2014). Additionally, among adults exposed to chronic psychological and physical violence during childhood, community cohesion has been found to be a protective factor against psychological distress (Greenfield & Marks, 2010). However, research on community cohesion among P-IPV exposed individuals is lacking. Understanding a potential mechanism as to why community factors may boost resilience may inform coordinated community responses for individuals who have experienced P-IPV (e.g., smaller tight knit communities with less access to formal IPV services may harness the power of community support by increasing the number of community events to increase community cohesion). The present study will thus add to the limited literature of the potential role community cohesion may play in the lives of those who have been exposed to P-IPV as children. Spirituality and Social Support The literature on spirituality in IPV exposed individuals is lacking. Studies examining spirituality in IPV survivors has shown that the built-in supportive network of people with shared beliefs and values improves health and functioning in adults (Howell & Miller-Graff, 2014). Previous research has also demonstrated that higher spirituality among women survivors of IPV 24 is positively associated with increased resilience (Howell et al., 2018). In addition, spirituality-focused clinical interventions and clinical interventions that include spiritual components have been linked to more positive health outcomes for children (Nouhi et al., 2017). The data being used in the current study focuses on the spirituality of the P-IPV exposed individuals and not the spirituality of their mothers. However, it is hypothesized that spirituality will have a similar positive relationship with resilience for P-IPV exposed individuals. As alluded to, this study examined the potential influence that spirituality had on adults who were exposed to parental IPV as children, Study Objectives The goal of this study is to investigate the role of and relations among individual, interpersonal, community/organizational factors, and resilience for adults exposed to IPV during childhood. The results will be interpreted using a social-ecological model of resilience for individuals exposed to IPV as children. Understanding the mechanisms that promote resilience in P-IPV exposed children can inform the creation of target driven interventions that may bolster resilience in adulthood for P-IPV exposed children. Research Questions (RQs). For adults exposed to P-IPV during childhood: 1) Does having a higher level of social support correspond with increased resilience? 2) Does having a higher level of spirituality correspond with increased resilience? 3) Does having a higher level of community cohesion correspond with increased resilience? 4) When accounting for the relationships between the predictor variables (i.e., between social support and community cohesion; social support and spirituality; 25 community cohesion and spirituality) does social support, spirituality, and community cohesion each uniquely correspond with resilience? Hypotheses. I hypothesize that for adults exposed to IPV during childhood: 1) There will be a significant direct positive relationship between social support and resilience. 2) There will be a significant direct positive relationship between spirituality and resilience. 3) There will be a significant direct positive relationship between community cohesion and resilience. 4) When accounting for the correlations between the predictor variables, there will be direct positive relationships between social support and resilience; spirituality and resilience; and community cohesion and resilience. CHAPTER II: METHODOLOGY Participants The current study will examine data from a subsample of the 22,412 Canadians aged 15 years and older living in a Canadian province or territory who completed the 2019 General Social Survey (GSS; Victimization) conducted by Statistics Canada. These participants were selected using simple random sampling without replacement of records (list of telephone numbers available to Statistics Canada made available by the Census and telephone companies) to select one member within each household. This survey is representative of the Canadian population and had a response rate of 37.6%. 26 The subsample that was analyzed included participants who reported exposure to parental intimate partner violence before the age of 15 years (n = 2278 participants, which was 10.2% of the total sample). Given the relationship between sample size and model complexity in structural equation modeling (SEM), it is recommended that sample size be decided using the ratio of number of participants (N) to number of parameters to estimate (q) ratio (Pituch & Stevens, 2015). Jackson (2003) recommended an optimal ratio of 20:1 for increased confidence in the results. The large sample size of this dataset made it ideal to employ statistical techniques to examine the interrelationships among protective factors for resilience in individuals exposed to P-IPV as children (using SEM). Sixty percent of the sample were females (n = 1,367), 39.6% were males (n = 901) and 0.4% (n = 10) were gender diverse. The age of the sample ranged from 15 to 97 years (Mage = 52.6; SDage = 15.6). The highest level of education varied amongst the sample from having completed less than a high school diploma or equivalent to a university degree above the bachelor’s level (see Table 2 for details). Table 2 Highest Education Level of Respondents Highest Education Level Achieved Number of respondents (%) Less than a high school diploma or its equivalent 345 (15.1%) High school diploma or a high school equivalency certificate 537 (23.6%) Trade certificate or diploma 264 (11.6%) College, CEGEP, or other non-university certificate/diploma (not trades) 490 (21.5%) University certificate or diploma below the bachelor level 102 (4.5%) Bachelor’s degree 351 (15.4%) University certificate, diploma, or degree above the bachelor’s level 189 (8.3%) 27 Most of the sample (69.6%; n = 1585) identified as White, 20.5% (n = 467) were Aboriginal, and 9.9% (n = 226) were visible minorities. Visible minority is defined based on the definition used in the GSS which refers to whether a person belongs to a visible minority group as defined by the Employment Equity Act and, if so, the visible minority group to which the person belongs. The Employment Equity Act defines visible minority as “persons other than Aboriginal peoples, who are non-Caucasian in race or non-white in colour.” For more information on respondent ethnicities, see Table 3. Please note that to protect the identities of the respondents, numbers less than 10 have been redacted and replaced with “<10” and corresponding percentages are not reported. Similarly, for respondents who reported more than one ethnicity, these data has been collapsed into one category “Multiple Ethnicities,” to protect their anonymity. See Table 4 for Aboriginal identities of the respondents. Table 3 Ethnicities of Respondents Ethnicity Number of Respondents (%) White 1585 (69.6%) Aboriginal 467 (20.5%) Chinese 62 (2.7%) South Asian 39 (1.7%) Black 36 (1.6%) Filipino 31 (1.4%) Latin American 30 (1.3%) *Multiple ethnicities 22 (1%) Arab < 10 Southeast Asian < 10 West Asian < 10 Korean < 10 Japanese < 10 Other < 10 Note. Multiple Ethnicities does not include multiple Aboriginal Ethnicities. 1 Table 4 Aboriginal Identities of Respondents Ethnicity Number of Respondents First Nations (North American Indian) 241 Metis 115 Inuk Multiple Aboriginal Identities 117 < 10 Most (82.9%; n = 1889) of the participants were born in Canada. In terms of citizenship, 87.4% (n = 1990) of the participants held Canadian citizenship, 8.3% (n = 190) had Canadian citizenship and citizenship with another country, and 4.3% (n = 97) had citizenship in a country other than Canada. One participant did not provide information on their citizenship. Of the sample, 70.5% (n = 1,607) reported that English was their primary language, 8.8% (n = 200) reported French as their primarily language, 19.8% (n =451) reported that a language other than English or French was their primary language, and less than one percent (n = 20) of the sample did not provide information on their primary language. Analyzing data from a randomly selected sample from across Canada that is representative of the Canadian population strengthens external validity (i.e., generalizability) of the results. This is particularly crucial when studying P-IPV because only a small proportion of IPV cases are ever reported to the police or brought to court (Public Health Agency of Canada, 2018). The utility of the analysis of these data was to create a model of resilience for individuals exposed to P-IPV as children that will be translated for the general population to help inform supports and programming for this population. In addition, most of the research that has been conducted on the relationships of various protective factors to promote resilience in P-IPV exposed individuals have been conducted in the United States and may not generalize to Canadians. 2 Design As previously mentioned, the data will be from the 2019 GSS on victimization, which were collected between April 15th, 2019 to March 31st, 2020 (data collection ceased due to the outbreak of COVID-19). One data collection method was via phone calls with a Statistics Canada interviewer who received sensitivity training to manage their own reactions to sensitive information and to provide mental health resources to respondents. Another method was an online survey that 60% of respondents completed. Additionally, in the territories, in person interviews were conducted. The survey included both Likert-style questions as well as open textbox questions that were coded by the GSS and Stats Can harmonized content program. It is important to note that the survey respondents could choose to complete the survey in either official language (i.e., English or French). Given the sensitive nature of the data, there may be underreporting of exposure to parental IPV for the interviews conducted in person or via a phone call. However, collecting data using an online questionnaire often produces a lower risk of response bias given that it is a more sensitive and private way to respond to the survey items. Given that 60% of the respondents completed the online survey, the impact of response bias may be minimal in this survey. Additionally, an important strength of these data is that, due to the length and breadth of the survey, there are multiple questions that can be combined to create a latent construct (e.g., community cohesion as a latent construct by combining neighbour helpfulness, neighbour trust, and neighbour connection). Measures Inclusion Criterion As previously mentioned, a subsample of the total survey data (i.e., select survey questions) will be analyzed. One question that assessed exposure to physical P-IPV before age 15 was used 3 to select the subsample of participants who were eligible for the current study. That is, those who provided responses that were greater than 0 to the following question will be included in the current sample: “Before age 15, how many times did you see or hear any one of your parents, step-parents or guardians do any of the following? Hit each other.” Responses greater than 0 were coded as physical P-IPV exposure in childhood. In addition, it has been established that psychological/emotional P-IPV exposure has significant negative impacts on children (Fritz & Roy, 2022). However, as the GSS does not have a clear measure of exposure to psychological/emotional P-IPV, this study will focus solely on physical P-IPV exposure. See Table 5 for variables with corresponding survey items, possible responses, and notes from the survey that pertain to each question (if applicable). The use of the 2019 GSS Victimization data allowed for a more fulsome measure of the variables used in this study as it is the only national self-report survey of victimization that provides data for the entire country (i.e., all provinces and territories). Independent Variables The following is a list of variables from the GSS (see Table 5) loosely organized based on the three levels from SAMSHA’s social-ecological model that has been used to conceptualize this study: individual (spirituality); interpersonal (informal social support); community and organizational (community cohesion). Of note, many of these variables can be viewed as belonging to more than one level of the socio-ecological model. One such variable that could exist at many levels of the socio-ecological model is spirituality. Given that spirituality encompasses a person’s individual relationship with a spiritual connection and has been conceptualized as an individual-level variable in previous research (see Howell et al., 2018), spirituality will be considered to be at the individual level in the current study. 4 Table 5 Relevant Variables from the 2019 General Social Survey Manifest variable Survey question Answer categories Latent variable Level I – Individual variables Religious identity “Religion of Respondent” No religion; Roman Catholic; United Church; Anglican; Presbyterian; Lutheran; Baptist; Eastern Orthodox; Jewish; Muslim; Buddhist; Hindu; Sikh; Jehovah’s Witness; Pentecostal; Ukrainian Catholic; Other; Valid skip; Don’t know; Refusal; Not stated. Spirituality Frequency of formal practice “Not counting events such as weddings or funerals, during the past 12 months, how often did you participate in religious activities or attend religious services or meetings?” At least once a week; At least once a month; At least 3 times a year; Once or twice a year; Not at all; Valid skip; Don’t know; Refusal; Not stated. Spirituality Frequency of informal practice “In the past 12 months, how often did you engage in religious or spiritual activities on your own?” At least once a week; At least once a month; At least 3 times a year; Once or twice a year; Not at all; Valid skip; Don’t know; Refusal; Not stated. Spirituality Importance of religion/spirituality “How important are your religious or spiritual beliefs to the way you live your life?” Very important; Somewhat important; Not very important; Not at all important; Valid skip; Don’t know; Refusal; Not stated. Spirituality Level II – Interpersonal variables P-IPV exposure during childhood “Before age 15, how many times did you see or hear any one of your parents, step-parents or guardians do any of the following? Hit each other”. Never; 1 or 2 times; 3 to 5 times; 6 to 10 times; More than 10 times; Valid skip; Don’t know; Refusal; Not stated. Close relationships “Approximately. How many relatives and friends do you have who you feel close to, that is, who you feel at each with, can talk to about what is on your mind?” None; one or two; three to five; six to nine; ten or more; Valid skip; Don’t know; Refusal; Not stated Social support Family trust “Using a scale of 1 to 5, where 1 means ‘Cannot be trusted at all’ and 5 means ‘Can be trusted a lot’, what is your level of trust in 1-Cannot be trusted at all; 2; 3; 4; 5- Can be trusted a lot; Valid skip; Don’t know; Refusal; Not stated.. Social support 5 each of the following groups of people? People in your family” Community belonging “How would you describe your sense of belonging to your local community?” Very strong; Somewhat strong; Somewhat weak; Very weak; No opinion; Valid skip; Don’t know; Refusal; Not stated Social support Level III – Community and organization variables Neighbour trust “(Using a scale of 1 to 5, where 1 means ‘Cannot be trusted at all’ and 5 means ‘Can be trusted a lot’, what is your level of trust in each of the following groups of people?) People in your neighbourhood” 1 – Cannot be trusted at all; 2; 3; 4; 5 – can be trusted a lot; Valid skip; Don’t know; Refusal; Not stated. Community cohesion Neighbour connection “Of the people in your neighbourhood, how many do you know?” Most of the people; Many of the people; A few of the people; None of the people; Valid skip; Don’t know; Refusal; Not stated. Community cohesion Neighbourhood helpfulness “Would you say this neighbourhood is a place where neighbours help each other?” Yes; No; Valid skip; Don’t know; Refusal; Not Stated. Community cohesion Resilience variables Mental health “In general, how is your mental health?” Excellent; Very good; Good; Fair; Poor; Valid skip; Don’t know; Refusal; Not stated. Resilience Physical health “In general, how is your health?” Excellent; Very good; Good; Fair; Poor; Valid skip; Don’t know; Refusal; Not stated. Resilience Life satisfaction “Using a scale of 0 to 10 where 0 means ‘very dissatisfied’ and 10 means ‘very satisfied’, how do you feel about your life as a whole right now?” 0 – very dissatisfied; 1; 2; 3; 4; 5; 6; 7; 8; 9; 10 – very satisfied; Valid skip; Don’t know; Refusal; Not stated. Resilience 6 Outcome Variable: The Conceptualization of Resilience for the Study As a reminder, for resilience the traumatic that was considered was the retrospective reporting of exposure to physical P-IPV before age 15. Second, a measure of participants’ subsequent psychological functioning and/or self-reported mental health will be used. Consistent with my conceptualization of resilience, the following variables will be included to assess resilience: subjective rating of mental health (labelled “Mental Health”), subjective rating of their health (labelled “General Health”), and subjective life satisfaction (labelled “Life Satisfaction”). For more information on survey variables, see Table 5. Data Analysis Data access was granted through the Statistics Canada Research Data Centre (RDC) at the University of Windsor. Data were analyzed using the lavaan package (0.6-18; Rosseel, 2012) in R version 4.1.3. Please note that the lavaan package uses the variance-covariance matrix for the data. For the analyses, all appropriate weights and Statistics Canada RDC conventions (e.g., rounding to a base of 10, meeting minimum cell counts of 15 for person or household variables and 40 for victimization incidents) were followed. For the counts of the demographic variables (e.g., religious affiliation) cell suppression was used by replacing values less than 10 with "<10.” This has helped to ensure anonymity of the survey respondents, making it harder to connect data to an individual. Structural Equation Modelling Prior to conducting the structural equation modelling (SEM), I conducted descriptive statistics for the indicators used in the measurement model (frequencies and percentages of responses). I used SEM to examine the research questions. Given the complex relationships previously identified among a multitude of factors that impact the outcomes of IPV, SEM 7 allowed for the examination of direct and correlational relationships between the variables with the incorporation of measurement error. The specific type of SEM that I used was structural regression (SR) modelling. In SR models, every variable is latent (i.e., a construct) with multiple indicators (i.e., measured variables). Similar to most SEM model types, SR models assume that the exogenous (or independent) variables do not have measurement error, whereas the endogenous (or dependent) variables are assumed to have measurement error (Kline, 2015). In SR models there are two components: the structural component, which includes the direct and indirect effects between the latent or observed variables; and the measurement component which is the association between the latent variables and their indicators (i.e., the manifest or observed variables). Original Model Specification The initial models were conceptualized (or specified) based on previous literature and theory and are added to the document below as appendices (see Appendix A – D). As a note, it was proposed that latent variable scaling would be used in which all factor loadings between the manifest variables and their respective latent variable were set to “unity” or one (Davvetas et al., 2020). This choice was made given the strong theoretical justification that all indicators measure the latent construct equally and directly. Given that the error or disturbance (i.e., the unexplained or residual variance) of latent variables were unscaled, the disturbances also received a scaling constant of one. All models analyzed the unstandardized solution (i.e., the covariance matrix). The first model was created to determine if the proposed protective factor variable, social support, was significantly related to resilience. In a similar vein, the second model was created to determine if the proposed protective factor, spirituality, was significantly related to resilience. The third model was created to determine if the proposed protective factor, community cohesion, 8 was significantly related to resilience. The final overall model was created to determine if the proposed protective factors of social support, spirituality, community cohesion, and resilience, respectively, were all significantly related. Given the complexity of spirituality, social support, community cohesion, and resilience, these topics were best operationalized as constructs. All four models were fully latent SR models that include three latent variables (i.e., constructs): Model 1: Social support and resilience (see Appendix A); Model 2: Spirituality and resilience (see Appendix B); Model 3: Community cohesion and resilience (see Appendix C). Model 4: Social support, spirituality, community cohesion and resilience (see Appendix D). As social support, spirituality, community cohesion, and resilience are all constructs or latent variables, all models are reflective measurement models in which the indicators are affected by the latent variable and are interchangeable. In essence, the latent variable is caused by its observed variables in an L –> M (latent to manifest) block as the latent construct exists independent of the measures (even if one indicator was deleted, the latent variable would still exist). According to Kline (2015), latent constructs cause the measured variables so the error that exists in these models is due to an inability to fully explain these measures – taking measurement error into account; error terms in indicators can be identified. It is required that reflective models have highly correlated indicators or measured variables. When there is a change in the construct, there will be a change in the manifest variables. Based on SAMSHA’s socio-ecological framework (2014), social support was conceptualized as belonging to the interpersonal level of the framework. Social support was measured using the indicator variables of close relationships, community belonging, and family trust, respectively. Next, the latent variable of resilience was measured using mental health, general health, and life satisfaction as indicators. It was proposed that spirituality, which is at the 9 individual level of the socio-ecological model, would be measured by religious identity, frequency of formal practice, importance of religion, and frequency of informal practice. Finally, community cohesion (from the community and organizational level of the socio-ecological model) was measured using neighbour trust, neighbour helpfulness, and neighbour connection (see Table 5 for additional information on survey items). Each model was evaluated to determine if it was overidentified. Overidentification is a requirement of SEM that states that the degrees of freedom must be greater than zero (dfm ≥ 0; calculated by pieces of information – parameter estimates). Essentially, there must be more pieces of information (nonredundant observations; p* from p* = p (p+1) /2 formula) than parameter estimates (q; variances of exogenous variables with error/disturbances, direct effects, and covariances). For SR models, there are additional requirements for identification which are: there must be at least two indicators (or measured/observable variables) per latent variable and the model must be recursive (i.e., there must be no bi-directional effects proposed between the endogenous variables in the model). The information regarding identification is provided under the figure for each corresponding model in the appendices. SEM Assumptions First, I assessed the assumptions of SEM. The first assumption is that the model is correctly specified a priori. That is, it is important to create models by specifying relationships among predictors based on past theory/research to reduce potential misspecification errors. However, this assumption is essentially always violated as there is no way to know if the proposed models are correct before testing them (Pituch & Stevens, 2015). Second, given the relationship between sample size and model complexity, it is recommended that sample size be decided using the ratio of number of participants (N) to number of parameters to estimate (q) 10 ratio (Pituch & Stevens, 2015). Jackson (2003) recommended an optimal ratio of 20:1 for increased confidence in the results. Third, I examined multicollinearity within the data by determining that there were no correlations greater than r = .9 between variables. In addition, variance inflation (VIF) values should be less than 6.7, and tolerance values should be greater than .15. Fourth, I evaluated multivariate or univariate normality. Specifically, I calculated Mardia’s coefficient for skewness and kurtosis to determine if the assumption of multivariate normality had been violated. The Shapiro Wilk values were used to assess univariate normality. However, given that the analyses were conducted with either maximum likelihood robust (MLR) correction to scale the chi-square statistic using the Yuan-Bentler correct for non-normality or weighted least squares mean and variance (WLSMV), this was not an issue for the analyses (Pituch & Stevens, 2015). Fifth, I investigated whether there are any outliers for the endogenous variables. That is, I examined Mahalanobis distance scores with a chi-square distribution on the initial dataset using a cut-off of p < .0001. Sixth, I assumed independence of observations as survey respondents should not have influenced each other’s responses given that the survey was completed by one member in each household. Seventh, I assessed the extent to which there were missing data. Finally, I assessed whether there were correlations between error terms. In SEM, it is assumed that there are no correlations between error terms—unless otherwise stated by the researcher in the conceptual model. Correlations Next, I conducted Spearman’s rank correlation analyses to examine the correlations between the indicator variables and their respective latent variables. The strength and direction of the correlations between the potential indicator variables and their respective latent variables were assessed. Redundancy of indicator variables was assessed. Indicator variables that are too 11 highly correlated (r > .9) are likely assessing the same construct, and thus the variables would have been removed. However, this was not the case for the current analyses. Examining Model Fit Once the models were refined, I analyzed the updated structural equation models. For the latent variables, error variances and lambda values (i.e., the latent factor loadings on the observed scores) were calculated and are reported below. I analyzed the parameter estimates to determine model fit, including the chi-square test (X 2; Jöreskog & Sörbom, 1989) with corresponding p values, standardized root mean square residual (SRMR; (Browne & Cudeck, 1989), root mean square error of approximation (RMSEA; Brown & Cudeck, 1989) with corresponding 90 percent confidence interval and p values, the comparative fit index (CFI; Bentler, 1990), the non-normed fit index (NNFI) also known as the Tucker Lewis Index (TLI; Bentler, 1990; Bentler & Bonett, 1980), and Akaike's (1987) Information Criteria (AIC). The following cut-offs were used to determine good model fit: X 2 with p > .05, SRMR ≤ .06, RMSEA ≤ .06, CFI ≥ .95, and TLI >.90 (Hu & Bentler, 1999). As the assumption of normality was violated, corrections were made based on the manifest variables. For the models examining the first research question, maximum likelihood robust (MLR) estimation was used with robust (Huber-White) standard errors and a scaled test statistic that is asymptotically equal to the Yuan-Bentler test statistic which corrects for non-normality and nonindependence of observations (Rosseel, 2012). For research question three and four, the parameter estimates were calculated using weighted least squares mean and variance (WLSMV) adjusted estimation due to the dichotomous nature of the neighbour connection variable (i.e., the response options were yes or no). The WLSMV estimation method is regarded as the best estimation method for binary or ordinal variables with few categories (AKA diagonal weighted least squares; DWLS; Newsom, 12 2023). WLSMV calculates a matrix of polychoric correlations which estimates what the association would be between the dichotomous variables if they were continuous and normally distributed. These polychoric correlations (aka latent correlations) are the correlations between the unobserved underlying continuous variables. The polychoric correlations are then used to create an asymptotic covariance matrix that acts as the weight matrix for the weighted least squares (WLS) estimation. The calculated path estimates represent the change in the latent correlations using the z scale (i.e., are standardized) for each unit change in the predictor. These path estimates are the equivalent to the values obtained using a probit regression (Newsom, 2023). Squared multiple correlations (R2) were calculated for the endogenous variables to determine how much the endogenous variables were explained by the regressions in the model. An R2 value of ≤ .20 suggests that the endogenous variable was not adequately explained by the regression(s) in the model and should be considered for removal from the model (Hooper et al., 2008). CHAPTER 3 RESULTS Data Recoding and Cleaning First, I recoded the responses of “valid skip,” “don’t know,” “refusal,” and “not stated” as missing data. In addition, I also recoded the response of “no opinion” for the sense of belonging to local community variable as missing. The sense of belonging to local community, neighbour connection, neighbour helpfulness, physical health, mental health, importance of religion/spirituality, and frequency of formal religious practice variables were reverse coded. Due to not having access to variables in the dataset (i.e., police IPV disclosure, friend IPV disclosure, family IPV disclosure, and neighbour IPV disclosure) that were initially proposed to construct the latent variable of social support, the latent variable of social support was 13 reconceptualized. The final latent variable of social support included the manifest variables of close relationships, community belonging, and family trust, respectively (see Table 5 for corresponding survey items). As the social support latent variable was updated and the spirituality latent variable had to be removed due to no longer meeting the identification criteria of having greater than two manifest variables, each model was reconceptualized. Next, I conducted a missing data analysis to determine if there were missing data patterns. The missing data pattern was examined by creating a missing data visualization and calculating a missing data summary (percentage of missing data for each variable, the proportion of cases missing for each variable, and the average number of missing values per case). The initial missing data analysis indicated that the frequency of informal religious/spiritual practice variable was missing responses from roughly half of the respondents. As a large proportion of data was missing for this variable, these data may have been MNAR (i.e., non-ignorable). As such, the frequency of informal religious/spiritual practice variable was removed from the analyses. In addition, the religious identity variable was also removed due to this variable being a nominal variable. As this latent variable no longer met the criteria of having greater than two manifest variables, the spirituality latent variable was not included in the current analyses and the models were reconceptualized. Please note that the removal of this latent variable led to a revision in the fourth research question that is stated in the results and discussion sections below. After the removal of the frequency of formal religious practice variable, a subsequent missing data analysis was conducted. There were 409 incomplete cases in the data, mean missing data = .006, the average missing data per case was .21, and the most missing data per case was seven data points. The variable with the highest percentage of missing data were the community belonging 14 (12.4%) and neighbour helpfulness (2.5%) variables. See Figure 1 for information on percentage of missing data for each variable. Figure 1 Bar Graph Displaying the Percentage of Missing Data for Each Variable Note. EHG_01 = education level; ISL_100 = close relationships; QIN_10 = neighbour connection ; QIN_20 = neighbour helpfulness ; recid = respondent identification number; REL_02 = frequency of formal religious practice; RELIG17H = religion of respondent; RLR_100 = importance of religion/spirituality; SBL_100 = community belonging; SLM_01 = life satisfaction ; SRH_110 = general health; SRH_115 = mental health ; TIP_10 = family trust; TIP_15 = neighbour trust. Results from the missing value analysis (MVA) revealed a significant Little MCAR test: (χ2 (1395, N = 2278) = 1502, p = .020). The significant result indicated that the data were not MCAR; the data were either MAR or nonignorable (Little, 1988). Based on the absence of patterns in the missing data visualization and summary, the data were determined to be MAR. Additional evidence to assume the data were MAR include the small proportion of missing data 15 combined with the knowledge that, under MNAR, very extreme scenarios would have to happen for your results to be overturned (also known as the tipping point analysis; SAS inc., 2019). Given that the missing data were assumed to be MAR, full information maximum likelihood (FIML) was used to handle missing data for the first research question. Due to the dichotomous nature of the neighbour connection variable used as a manifest variable in the models for research questions three and four, WLSMV was the chosen estimation method. WLSMV handles missing data using a partially pairwise deletion estimation that is not considered a full information estimation approach (Newsom, 2023). Given the small ratio of missing data for the current dataset, partially pairwise deletion was an appropriate method to handle missing data (Newsom, 2023). Outliers for the endogenous variables were identified using Mahalanobis squared distance scores (χ2 = 77.80; alpha = .0001; 99.99% CI). There were 312 outliers. In an attempt to retain as much of the original data as possible and given that Statistics Canada had already followed a rigorous data validation process to reduce errors and bias, these data points were not removed (Government of Canada, 2022). Due to missing demographic information (i.e., gender, age, or ethnicity) 62 participants were excluded. This left an overall sample of N = 2,278. A flow diagram depicting the data cleaning procedure can be found below: 16 Figure 2 Data Cleaning Procedures Assumption Testing The assumption of the absence of multicollinearity was met as the only correlations within the data that were greater than r = .9 were ethnicity variables. In addition, the variance inflation (VIF) values were less than 6.7, and the tolerance values were greater than .15. Multivariate normality was assessed using (a) Mardia’s coefficient of skewness and kurtosis (Mardia, 1970), (b) univariate skewness and kurtosis statistics (skewness is <-3 or > +3; Tabachnik & Fidell, 2013), and (c) visual inspection of the generated distributions. Mardia’s coefficient for multivariate skewness (Mardia’s coefficient = 765739.7, p < .001) and kurtosis (Mardia’s coefficient = 1190.6, p <.001) were calculated on the cleaned dataset using the MVN package in R (Korkmaz et al., 2014). Both Mardia’s skewness and kurtosis values were significant, indicating that the assumption of multivariate normality was violated. Based on the Shapiro-Wilk values (see Table 6) and QQ Plot (see Figure 3) univariate normality was also violated. These N = 2278 Reported all demographicInformationN = 22,412 Completed the 2019 GSS VictimizationN = 2340 Reported physical P-IPV exposureDid not report P-IPV exposure (n = 20,072)• Missing gender (n = 3)• Missing education (n = 17)• Missing Ethnicity (n = 42) 17 violations to univariate normality serve as further evidence that multivariate normality was violated (Tabachnik & Fidell, 2013). Table 6 Shapiro Wilk Normality Test Values Variable Shapiro-Wilk value (W) Age .99*** Exposure to P-IPV .77*** Close relationships .86*** Community belonging .85*** Family trust .57*** Neighbour trust .90*** Neighbour helpfulness .47*** Frequency of formal practice .71*** Importance of religion/spirituality .81*** General health .91*** Mental health .91*** Life satisfaction .85*** ***p < .001. 18 Figure 3 Q-Q Plot to Examine Univariate Normality Note. The normal q-q plot indicates that the data are right (positively) skewed. I calculated the correlations between the manifest variables and latent variables to examine the strength and direction of the relationships between the indicator variables and their respective latent variables for each model. Some correlations were weaker than the desired moderate to strong relationship (r < .4). Due to the archival nature of these data, there were no other available indicators that fit the constructs well. Given that SEM requires at least three indicators per latent variable, these indicator variables were retained. Based on the correlations, there was no redundancy of indicator variables (i.e., no indicators variables were too highly correlated r > .9). The number of instances of exposure to physical P-IPV reported by the sample is presented in Table 7. Of the respondents who reported exposure to physical P-IPV before age 15, roughly half of the respondents reported exposure to one to two episodes. 19 Table 7 Number of Physical P-IPV Exposures Reported by Respondents Number of reported P-IPV exposures n (%) 1 or 2 times 1001 (43.9%) 3 to 5 times 436 (19.1%) 6 to 10 times 201 (8.9%) More than 10 times 640 (28.1%) 20 Table 8 Frequencies of Responses for Survey Items Included in Structural Regression Models Variable responses Frequency (%) Close relationships None 3 (0.1%) One to two 461 (20.2%) Three to five 752 (33.0%) Six to nine 355 (15.6%) Ten or more 603 (26.5%) Missing response 104 (4.6%) Community belonging Very weak 223 (9.8%) Somewhat weak 382 (16.8%) Somewhat strong 881 (38.7%) Very strong 509 (22.3%) No opinion 275 (12.1%) Missing response 8 (0.4%) Family trust 1 – Cannot be trusted at all 42 (1.8%) 2 54 (2.4%) 3 163 (7.2%) 4 353 (15.5%) 5 – Can be trusted a lot 1657 (72.7%) Missing response 9 (0.4%) Neighbour trust 1 – Cannot be trusted at all 108 (4.7%) 2 244 (10.7%) 3 719 (31.6%) 4 696 (30.6%) 5 – Can be trusted a lot 496 (21.8%) Missing response 15 (0.7%) Neighbour connection None of the people 152 (6.7%) A few of the people 1186 (52.1%) Many of the people 385 (16.9%) Most of the people 551 (24.2%) 21 Missing response 4 (0.2%) Neighbour helpfulness No 406 (17.8%) Yes 1816 (79.7%) Missing response 56 (2.5%) General gealth Poor 145 (6.4%) Fair 364 (16.0%) Good 821 (36.0%) Very good 668 (29.3%) Excellent 276 (12.1%) Missing response 4 (0.2%) Mental gealth Poor 100 (4.4%) Fair 330 (14.5%) Good 759 (33.3%) Very good 696 (30.6%) Excellent 388 (17.0%) Missing response 5 (.2%) Life satisfaction 0 – Very Dissatisfied 39 (1.7%) 1 9 (0.4%) 2 33 (1.4%) 3 67 (2.9%) 4 69 (3.0%) 5 195 (8.6%) 6 177 (7.8%) 7 359 (15.8%) 8 565 (24.8%) 9 268 (11.8%) 10 – Very satisfied 485 (21.3%) Missing response 12 (0.5%) 22 Table 9 shows the bivariate Spearman’s correlations (ρ) between all the variables included in the proposed SEM models. Overall, the correlational analysis showed that all variables were significantly positively correlated. The correlations between the manifest variables that made up the latent variable of social support were weakly correlated (ρ = .15 – .20). The manifest variables that made up the latent variable of community cohesion were also weekly correlated ranging from ρ = .15 – .20. Finally, the manifest variables that made up the latent variable of community cohesion were moderately correlated (ρ = .51 – .52) except for life satisfaction and mental health which were weakly correlated (ρ = .39). 23 Table 9 Correlations Among Variables in Proposed Structural Models 1 2 3 4 5 6 7 8 9 Social support 1. Close relationships -- 2. Community belonging .21* -- 3. Family trust .20** .15** -- Community cohesion 4. Neighbour trust .22** .34** .31** -- 5. Neighbour connection .18** .44** .08** .29** -- 6. Neighbour helpfulness .15** .27** .13** .36** .25** -- Resilience 7. General health .14** .12** .15** .14** .07** .11** -- 8. Mental health .17** .19** .20** .20** .11** .13** .51** -- 9. Life satisfaction .24** .25** .19** .20** .16** .16** .52** .39** -- Note. The corresponding latent variable for each manifest variable is written in bold. *p < .05; **p < .01. 24 Research Question 1: Does Having a Higher Level of Social Support Correspond With Increased Resilience? Model 1R: Direct Relationship Between Social Support and Resilience As the social support variable was re-conceptualized, the new model examined the direct relationship between social support (measured using the variables of close relationships, community belonging and family trust) and resilience (measured using the mental health, physical health, and life satisfaction variables). 25 Figure 4 Model 1 Revised: A Structural Regression Model of a Direct Relationship Between Social Support and Resilience Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols: a circle = latent variable or construct, a rectangle = measured or observed variable, a circle with a D = a disturbance which is the error of a latent variable, a line with a single arrowhead (e.g., –>) = hypothesized directional causal effects or direct effects on endogenous variables, two-headed curved arrows that exit and enter the same variable = variance of an exogenous variable, and a larger curved line with two arrow heads between two variables = covariance. See Table 5 for corresponding survey questions for the observed variables. Mental Health Physical HealthLife SatisfactionD21ResilienceSocial SupportD1 1Close RelationshipsCommunity Belonging Family TrustE11E2 E31 1E41E5 E61 11 1 26 Model 1R Results. Using the formula p* = p (p+1) /2, there were 6(7)/ 2 = 21 nonredundant observations. There were 6 variances of exogenous variables and 7 direct effects for a total of 13 parameter estimates (q). The calculation is thus: 21 pieces of information – 13 parameter estimates = 8; dfm ≥ 0. The additional requirements for identification of SR models were also met (i.e., there are at least two indicators per latent variable and the model is recursive). As such, Model 1R was overidentified. Recall that for the latent factors, their variances were set to 1 and the means were set to 0. The parameter estimates were calculated using MLR estimation (i.e., the standard errors and fit indices were adjusted using the Yuan-Bentler correction for non-normality and nonindependence of observations); as such, the robust parameter estimates are reported. The chi-square statistic had a Yuan-Bentler scaling correction factor of 0.95. The Yuan-Bentler χ2 (8, N = 2278) = 46.68, p < .001 suggested the model did not fit the data well. It is standard practice in SEM to report the Chi-square goodness of fit test. However, the Chi-square goodness of fit test is sensitive to sample size. In cases, such as this current thesis, where the sample size is large, chi-square almost always rejects the null hypothesis, suggesting poor model fit (Hooper et al., 2008). Fit indices indicated a good fit (robust comparative fit index (R-CFI) = .96; robust Tucker Lewis index (R-TLI) = .92; standardized root means square residual (SRMR = .024); and the root mean square error of approximation (R-RMSEA) = .045; 90% CI [.033, .058]; p < .05). Overall, Model 1R fit the data well. Parameter estimates for the structural paths representing the direct effects on social support and resilience are presented in Table 10. The parameter estimates indicated that the factor loadings were significant and in the expected direction. For Model 1R, the R2 values for close relationships, community belonging, and life satisfaction, respectively, were less than .20 (see Table 11). 27 Table 10 Model 1R: Latent Variable Loadings and Variances Std est. Est. SE z p Latent variable loadings Social support -> Close relationships .32 0.41 0.05 8.52 <.001 Social support -> Community belonging .34 0.31 0.04 7.48 <.001 Social support -> Family trust .53 0.46 0.05 9.64 <.001 Resilience -> General health .64 0.68 0.04 16.61 <.001 Resilience -> Mental health .80 0.85 0.05 17.85 <.001 Resilience -> Life satisfaction .01 0.04 0.09 0.48 .632 Variances Resilience 1.00 1.00 -- -- -- Social support 1.00 1.00 -- -- -- Close relationships 0.90 1.41 0.04 32.09 <.001 Community belonging 0.89 0.77 0.03 24.52 <.001 Family trust 0.72 0.55 0.04 12.90 <.001 General health 0.60 9.02 0.18 50.19 <.001 Mental health 0.37 0.42 0.08 5.30 <.001 Life satisfaction 1.00 9.02 0.18 50.19 <.001 Note. Std. Est = standardized estimate (𝛽) for the latent variable loadings which shows the strength of relationships between the indicator and latent variable. Est = unstandardized estimate for the latent variable loadings (B) – the slope of the indicator variables with the latent variable. SE = standard errors of each estimate; Latent variable loadings = the standardized loadings of manifest variables onto latent variables. Variances (AKA error variance or residuals) = the variances of the manifest and latent variables – how much variance of a given parameter does not contribute to the latent variable. 28 Table 11 Model 1R: Squared Multiple Correlation (R2) Values for Each Exogenous Variable Manifest variable R2 Social support .23 Close relationships .10 Community belonging .11 Family trust .28 General health .41 Mental health .64 Life satisfaction .00 An increase in social support significantly corresponded with an increase in resilience such that a one standard deviation increase in social support corresponded to a .52 standard deviation increase in resilience. The 𝛽 value of .52 (SE = 0.08, p < .001) indicated that there was a large statistically significant direct relationship between social support and resilience. Model 1B Association Between Social Support and Resilience I next tested an alternate model in which social support and resilience were allowed to covary (i.e., if social support and resilience were significantly correlated). This model allowed for the examination of the effects that resilience may have on social support given that individuals who have higher resilience may tend to seek out more social support. 29 Figure 5: Model 1B: A Structural Regression Model of the Correlational Relationship Between Social Support and Resilience Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols.Mental Health GeneralHealthLife Satisfaction1D21Resilience1 1Social SupportD1 11Close RelationshipsCommunity Belonging Family Trust1 1E11E2 E31 1E41E5 E61 1 30 Model 1B Results. Model 1B’s identification was identical to that of Model 1R. Again, I fixed the variances of the latent factors to 1 and means to 0 and used MLR with the Yuan-Bentler correction. Similar to Model 1R, all fit indices indicated a good fit, except for the chi-square analysis which was likely due to the large sample size, Yuan-Bentler χ2 (8, N = 2278) = 46.68, p < .001; R-CFI = .96; R-TLI = .93; SRMR = .024; and R-RMSEA = .045; 90% CI [.033, .058]; p < .05. The parameter estimates indicated that the factor loadings were significant and in the expected direction (see Table 12). There was a moderate positive statistically significant correlation between social support and resilience (r = .52, SE = .08, p < .001). The R2 values for close relationships, community belonging, and life satisfaction, respectively, were less than .20 (see Table 13). 31 Table 12 Model 1B: Latent Variable Loadings and Variances Indicators Std Est. Est. SE z p Latent variable loadings Social support -> Close relationships 0.32 0.41 0.05 8.52 <.001 Social support -> Community belonging 0.34 0.31 0.04 7.48 <.001 Social support -> Family trust 0.53 0.46 0.05 9.64 <.001 Resilience -> General health 0.64 0.68 0.04 16.61 <.001 Resilience -> Mental health 0.80 0.85 0.05 17.85 <.001 Resilience -> Life satisfaction 0.01 0.04 0.09 0.48 .632 Variances Resilience 1.00 1.00 -- -- -- Social support 1.00 1.00 -- -- -- Close relationships 0.90 1.41 0.04 32.09 <.001 Community belonging 0.89 0.77 0.03 24.52 <.001 Family trust 0.72 0.55 0.04 12.90 <.001 General health 0.60 9.02 0.18 50.19 <.001 Mental health 0.37 0.42 0.08 5.30 <.001 Life satisfaction 1.00 9.02 0.18 50.19 <.001 Note. Std. est = standardized estimate 𝛽. Est = unstandardized estimate B. Standard errors (SE) = standard errors of each estimate. 32 Table 13 Model 1B: Squared Multiple Correlation (R2) Values for Each Exogenous Variable Manifest variable R2 Social support .26 Close relationships .11 Community belonging .11 Family trust .28 General health .41 Mental health .64 Life satisfaction .00 Model 1C Direct Effect Between Social Support and Resilience with Fixed Factor Loadings Next, I examined a third model in which the lambda values λ (i.e., the latent factor loadings on the observed scores) were set to unity (or one). This decision was based on previous research suggesting equal relationships between each manifest variable and the latent variable. 33 Figure 6 Model 1C: A model of the Direct Relationship Between Social Support and Resilience with Fixed Factor Loadings Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols. Mental Health Physical HealthLife Satisfaction1D21Resilience1 1Social SupportD1 11Close RelationshipsCommunity Belonging Family Trust1 1E11E2 E31 1E41E5 E61 1 34 Model 1C Results. Using the formula p* = p (p+1) /2, there were 6(7)/ 2= 21 non-redundant observations. There were 8 variances of exogenous variables and 1 direct effect for a total of 9 parameter estimates (q). Given that 21 pieces of information – 9 parameter estimates = 12; dfm ≥ 0. The additional requirements for identification of SR models were also met. As such, Model 1C was overidentified. The parameter estimates were calculated using MLR estimation with the Yuan-Bentler correction. The chi-square statistic had a Yuan-Bentler scaling correction factor of 1.05. Fit indices indicated poor fit, Yuan-Bentler χ2 (12, N = 2278) = 128.80, p < .001; R-CFI= .87; R-TLI = .84; SRMR = .055; R-RMSEA = .065; 90% CI [.057, .078]; p < .05. Overall, Model 1C did not fit the data well. However, the parameter estimates indicated that the factor loadings were significant and in the expected direction (see Table 14). Interpretation of the parameter estimates was limited due to poor model fit. Parameter estimates are likely biased; thus, observed effects should be interpreted with caution. The R2 values for close relationships, community belonging, and life satisfaction, respectively, were less than .20 (see Table 15). An increase in social support significantly corresponded with increased resilience such that a one standard deviation increase in social support corresponded to a .56 standard deviation increase in resilience. The 𝛽 value = .56 (SE = 0.08, p < .001) indicated that there was a large statistically significant direct relationship between social support and resilience. 35 Table 14 Model 1C: Latent Variable Loadings and Variances Indicators Std Est. Est. SE z p Latent variable loadings Social support -> Close relationships .31 1.00 – – – Social support -> Community belonging .41 1.00 – – – Social support -> Family trust .45 1.00 – – – Resilience -> General health .69 1.00 – – – Resilience -> Mental health .69 1.00 – – – Resilience -> Life satisfaction .24 1.00 – – – Variances Close relationships .91 1.42 0.03 42.23 <.001 Community belonging .83 0.74 0.03 29.69 <.001 Family trust .80 0.60 0.04 16.33 <.001 General health .52 0.58 0.03 22.19 <.001 Mental health .52 0.58 0.03 20.37 <.001 Life satisfaction .95 9.13 0.20 45.19 <.001 Social support 1.00 0.15 0.02 8.29 <.001 Resilience 1.00 0.53 0.03 20.00 <.001 Note. Std. est = standardized estimate 𝛽. Est = unstandardized estimate B. Standard errors (SE) = standard errors of each estimate. 36 Table 15 Model 1C: R2 Values for Each Variable Manifest variable R2 Close relationships .10 Community belonging .17 Family trust .20 General health .48 Mental health .48 Life satisfaction .06 Social support .31 Comparison of Models for Research Question 1: The Relationship between Social Support and Resilience To determine which model best fit the data from three proposed models, the model fit information for each of the models was compared and reported in Table 16. In addition, information-based criteria measures were analyzed (see Table 17). The model with the smallest information criteria values is said to be the model that demonstrates highest predictive validity (in essence, the model would be more generalizable to subsequent samples from the population) compared to the other competing models (Pituch & Stevens, 2015). Previous research suggests that the Bayesian information criteria (BIC) is a superior measure to the Akaike information criteria (AIC) as it tends to select less complex models with fewer parameters (Whittaker & Stapleton, 2006). The AIC on the other hand tends to favour more complex models with a higher number of parameters (Bozdogan, 1987; Brown & Cudeck, 1989). Based on the fit statistics as well as the information criteria values, Model 1R and Model 1B were the best fitting models and were considered equivalent models. As such, allowing the factor loadings to be freely estimated resulted in a model with improved fit to the data. This suggested that the manifest variables do not equally predict the latent constructs in the model. 37 Table 16 Comparison of Model Fit Statistics for Model Examining Research Question 1 Model Yuan-Bentler χ2 df SRMR R-RMSEA (90% CI) R-CFI R-TLI 1R: Direct effect - freely estimated λ 46.68*** 8 .02 .05 (.30, .60)* .96 .92 1B: Correlation – freely estimated λ 46.68*** 8 .02 .05 (.30, .60)* .96 .92 1C: Direct effect – fixed λ 128.80*** 12 .06 .65 (.50, .80)* .87 .84 Note. λ = factor loadings. Model 1R examined the direct effect between social support and resilience with freely estimated factor loadings; Model 1B examined the correlation between social support and resilience with freely estimated factor loadings; Model 1C examined the direct effect between social support and resilience – fixed factor loadings. * p < .05. ** p < .01. *** p < .001. Table 17 Information Criteria Values for Models Examining Research Question 1 Models AIC BIC 1R: Direct effect - freely estimated λ 42618.43 42618.32 1B: Correlation – freely estimated λ 42618.43 42727.32 1C: Direct effect – fixed λ 42701.68 42787.65 Note. Model 1R examined the direct effect between social support and resilience with freely estimated factor loadings; Model 1B examined the correlation between social support and resilience with freely estimated factor loadings; Model 1C examined the direct effect between social support and resilience – fixed factor loadings. AIC = Akaike information criteria; BIC = Bayesian information criteria. 38 Question 2: Does Having a Higher Level of Spiritualty Correspond With Increased Resilience? Model 2: A Structural Regression Model of Spirituality and Resilience As previously mentioned, the frequency of informal religious/spiritual practice variable and the religious identity variables were removed from this model. As this latent variable no longer met the criteria of having greater than two manifest variables, the spirituality latent variable was not included in the current analyses and this model was unable to be analyzed. Research Question 3: Does Having a Higher Level of Community Cohesion Correspond With Increased Resilience? Model 3A: A Structural Regression Model of Community Cohesion and Resilience Model 3 examined the direct effect of community cohesion (manifest variables: neighbour trust, neighbour helpfulness, and neighbour connection) on resilience (manifest variables: mental health, physical health, and life satisfaction). 39 Figure 7 Model 3A: A Structural Regression Model of the Direct Relationship Between Community Cohesion and Resilience Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols. Mental Health GeneralHealthLife Satisfaction1D21Resilience1 1Community CohesionD1 11NeighbourTrustNeighbourHelpfulnessNeighbourConnection1 1E11E2 E31 1E41E5 E61 1 40 Model 3A Results. There were 21 non-redundant observations – 13 parameter estimates = 8; dfm ≥ 0. The additional requirements for identification of SR models were also met. As such, Model 3A was overidentified. Recall that for the latent factors, their variances were set to 1 with means set to 0. The parameter estimates were calculated using WLSMV estimation due to the dichotomous nature of the neighbour connection variable (i.e., the response options were yes or no). The WLSMV estimation method is regarded as the best estimation method for binary or ordinal variables with few categories (AKA diagonal weighted least squares; DWLS; Newson, 2023). As WLSMV estimator adjusts for non-normality, the robust parameter values are reported. The chi-square statistic had a scaling correction factor of 0.85. The scale corrected χ2 (8, N = 2278) = 18.93, p < .001 suggested the model did not fit the data well which is likely biased due to the large sample size. Fit indices indicated good fit (R-CFI = .99; R-TLI = .99; SRMR = .023; R-RMSEA = .025; 90% CI [.010, .040]; p < .05). Parameter estimates indicated that the factor loadings were significant and in the expected direction (see Table 18). For Model 3R, the R2 values for community cohesion and life satisfaction, respectively, were less than .20 (see Table 19). The unstandardized B value = .34 (SE = 0.04, p < .001) indicated there was a weak significant direct relationship between the latent variables of community cohesion and resilience. 41 Table 18 Model 3A: Latent Variable Loadings and Variances Variable/Effect Est. SE Z p Latent variable loadings Community cohesion -> Neighbour trust 0.66 0.03 24.07 <.001 Community cohesion -> Neighbour connection 0.48 0.02 19.68 <.001 Community cohesion -> Neighbour helpfulness 0.76 0.03 24.17 <.001 Resilience -> General health 0.86 0.04 19.87 <.001 Resilience -> Mental health 0.66 0.04 18.48 <.001 Resilience -> Life satisfaction 0.14 0.02 7.04 <.001 Variances Resilience 1.00 – – – Community cohesion 1.00 – – – Neighbour trust 0.52 – – – Neighbour connection 0.74 – – – Neighbour helpfulness 0.36 – – – General health 0.57 – – – Mental health 0.27 – – – Life satisfaction 0.98 – – – Note. Std. est = standardized estimate 𝛽. Est = unstandardized estimate B. Standard errors (SE) = standard errors of each estimate. 42 Table 19 Model 3A: R2 Values for Each Variable Manifest variable R2 Community cohesion .10 Neighbour trust .48 Neighbour connection .26 Neighbour helpfulness .64 General health .43 Mental health .73 Life satisfaction .02 Model 3B Association Between Community Cohesion and Resilience I next assessed a model to determine the correlation between community cohesion and resilience. This model allowed for the examination of the effects that resilience may have on community cohesion given that individuals who have higher resilience may tend to seek out more cohesive communities. 43 Figure 8 Model 3B: A Structural Regression Model of the Correlational Relationship Between Community Cohesion and Resilience Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols. Mental Health GeneralHealthLife SatisfactionD21ResilienceE41E5 E61 1Community CohesionD1 1NeighbourTrustNeighbourHelpfulnessNeighbourConnectionE11E2 E31 11 1 44 Identification. Model 3B’s identification was identical to model 3R. I once again fixed the variances of the latent factors to 1 and means to 0 and used WLSMV estimation. Similar to Model 3R, all fit indices indicated a good fit, except for the chi-square analysis which was likely due to the large sample size (Scale-corrected χ2 (8, N = 2278) = 18.93, p < .001; R-CFI = .99; R-TLI = .99; SRMR = .023; RMSEA = .025; 90% CI [.010, .040]; p < .05). The parameter estimates indicated that the factor loadings were significant and in the expected direction (see Table 20). There was a weak statistically significant correlation between the latent variables of social support and community cohesion (r = .32, SE = 0.03, p < .001). The R2 value for life satisfaction was less than .20 (see Table 21). 45 Table 20 Model 3B: Latent Variable Loadings and Variances Variable/Effect Est. SE Z p Latent variable loadings Community cohesion -> Neighbour trust 0.69 0.03 25.35 <.001 Community cohesion -> Neighbour connection 0.51 0.02 20.15 <.001 Community cohesion -> Neighbour helpfulness 0.80 0.03 25.03 <.001 Resilience -> General health 0.86 0.04 19.87 <.001 Resilience -> Mental health 0.66 0.04 18.48 <.001 Resilience -> Life satisfaction 0.14 0.02 7.04 <.001 Variances Resilience 1.00 – – – Community cohesion 1.00 – – – Neighbour trust 0.52 – – – Neighbour connection 0.74 – – – Neighbour helpfulness 0.36 – – – General health 0.57 – – – Mental health 0.27 – – – Life satisfaction 0.98 – – – Note. Std. est = standardized estimate 𝛽. Est = unstandardized estimate B. Standard errors (SE) = standard errors of each estimate. 46 Table 21 Model 3B: R2 Values for Each Variable Manifest variable R2 Neighbour trust .48 Neighbour connection .26 Neighbour helpfulness .64 General health .43 Mental health .73 Life satisfaction .02 47 Model 3C Direct Effect Between Community Cohesion and Resilience with Fixed Factor Loadings I examined a third model with the lambda values λ (i.e., the latent factor loadings on the observed scores) set to unity (or one). This decision was based on previous research suggesting equal relationships between each manifest variable and the latent variable. 48 Figure 9 Model 3C: A Structural Regression Model of Community Cohesion and Resilience with Fixed Factor Loadings Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols. Mental Health GeneralHealthLife Satisfaction1D21Resilience1 1Social SupportD1 11Close RelationshipsCommunity Belonging Family Trust1 1E11E2 E31 1E41E5 E61 1 49 Model 3C Results. There were 21 non-redundant observations – 9 parameter estimates = 12; dfm ≥ 0. The additional requirements for identification of SR models were also met. As such, Model 3C was overidentified. The parameter estimates were calculated using WLSMV estimation. Fit indices indicated poor fit (Scale corrected χ2 (12, N = 2278) = 249.39, p < .001; R-CFI = .76; R-TLI = .70; SRMR = .076; R-RMSEA = .095; 90% CI [.085, .106]; p < .05). Parameter estimates indicated that the factor loadings were significant and in the expected direction (see Table 22). Interpretation of the parameter estimates was limited due to poor model fit. The R2 values for neighbour trust, neighbour connection, life satisfaction, and community cohesion were less than .20 (see Table 23). There was a weak significant direct relationship between the latent variables of community cohesion and resilience (B = .14, SE = 0.02, p <.001). 50 Table 22 Model 3C: Latent Variable Loadings and Variances Variable/Effect Std. Est. Est. SE Z p Latent variable loadings Community cohesion -> Neighbour trust .33 1.00 – – – Community cohesion -> Neighbour connection .39 1.00 – – – Community cohesion -> Neighbour helpfulness .92 1.00 – – – Resilience -> General health .64 1.00 – – – Resilience -> Mental health .65 1.00 – – – Resilience -> Life satisfaction .23 1.00 – – – Variances Neighbour trust – 1.05 0.03 35.42 <.001 Neighbour connection – 0.73 0.02 41.78 <.001 Neighbour helpfulness – 0.02 0.01 3.65 <.001 General health – 0.67 0.03 24.54 <.001 Mental health – 0.65 0.03 22.10 <.001 Life satisfaction – 8.55 0.18 46.89 <.001 Community cohesion – 0.13 0.01 16.85 <.001 Resilience – 0.47 0.03 17.66 <.001 Note. Std. est = standardized estimate 𝛽. Est = unstandardized estimate B. Standard errors (SE) = standard errors of each estimate. 51 Table 23 Model 3C: R2 Values for Each Variable Manifest variable R2 Neighbour trust .11 Neighbour connection .15 Neighbour helpfulness .84 General health .42 Mental health .41 Life satisfaction .05 Community cohesion .07 Comparison of Models for Research Question 3: The Relationship between Community Cohesion and Resilience To determine which model best fit the data from three proposed models, the model fit information for each of the models was compared and reported in Table 24. As this model used WLSMV estimation, the information criteria values were unable to be calculated. Based on the fit statistics, Model 3A and Model 3B were the best fitting models and were considered equivalent models. As such, allowing the factor loadings to be freely estimated resulted in a model with improved fit to the data. This suggested that the manifest variables do not equally predict the latent constructs in the model. Table 24 Comparison of Model Fit Statistics for Models Examining Research Question 3 Model Yuan-Bentler χ2 df SRMR R-RMSEA (90% CI) R-CFI R-TLI 3A: Direct effect - freely estimated λ 18.93*** 8 .02 0.25 (.10, .40)* .99 .99 3B: Correlation – freely estimated λ 18.93*** 8 .02 0.25 (.10, .40)* .99 .99 52 3C: Direct effect – fixed λ 249.39*** 12 .08 0.95 (.90, 1.10 )* .76 .70 Note. λ = factor loadings. Model 3A examined the direct effect between community cohesion and resilience with freely estimated factor loadings; Model 3B examined the correlation between community cohesion and resilience with freely estimated factor loadings; Model 3C examined the direct effect between community cohesion and resilience – fixed factor loadings. * p < .05. *** p < .001. Research Question 4: When Accounting for the Relationship Between Social Support and Community Cohesion, Does an Increase in Social Support and Community Cohesion Each Uniquely Correspond With Increased Resilience? Model 4R: Social Support, Community Cohesion, and Resilience As the spirituality latent variable had been removed from analyses, the new Model 4 examined the correlation between the latent variables of community cohesion (manifest variables: neighbour trust, neighbour helpfulness, and neighbour connection) and social support (manifest variables: close relationships, community belonging and family trust), as well as the direct relationships between social support and resilience (manifest variables: mental health, physical health and life satisfaction variables) and community cohesion and resilience. 53 Figure 10 Model 4R: A Structural Regression Model of Community Cohesion, Resilience, and Social Support Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols. Community CohesionD1 1NeighbourTrustNeighbourHelpfulnessNeighbourConnectionE11E2 E31 1Close RelationshipsCommunity BelongingFamily TrustD21Social SupportE41E5 E61 1ResilienceMental Health General Health Life SatisfactionE71E8 E31 1D311 1 11 1 1111 54 Model 4R Results. There were 45 pieces of information – 15 parameter estimates = 30; dfm ≥ 0. The additional requirements for identification of SR models were also met. As such, Model 4R was overidentified. Here, the lambda values λ (i.e., the latent factor loadings on the observed scores) were set to unity (or one). The parameter estimates were calculated using WLSMV estimation. Fit indices indicated poor fit, Yuan Bentler χ2 (30, N = 2278) = 724.86, p < .001; R-CFI = .59; R-TLI = .51; SRMR = .099; R MSEA = .110; 90% CI [.103, .117]; p < .05. Parameter estimates indicated that the factor loadings were significant and in the expected direction (see Table 25). Interpretation of the parameter estimates was limited due to poor model fit. Parameter estimates are likely biased; thus, observed effects should be interpreted with caution. A statistically significant moderate positive relationship was found between the latent variables of social support and resilience (B = .58, SE = 0.02, p <.001). Next, there was a weak statistically significant positive relationship between community cohesion and resilience (B = .19, SE = 0.01, p < .001). Finally, there was a strong correlation between social support and community cohesion (r = .79, SE = .10, p <.001). The R2 values were less than .20 for close relationships and life satisfaction (see Table 26). 55 Table 25 Model 4R: Latent Variable Loadings and Variances Std. Est Est SE z p Latent variable loadings Social support -> Close relationships 0.30 1.00 – – – Social support -> Community belonging 0.40 1.00 – – – Social support -> Family trust 0.43 1.00 – – – Resilience -> General health 0.64 1.00 – – – Resilience -> Mental health 0.65 1.00 – – – Resilience -> Life satisfaction 0.23 1.00 – – – Community cohesion -> Neighbour trust 0.33 1.00 – – – Community cohesion -> Neighbour connection 0.38 1.00 – – – Community cohesion -> Neighbour helpfulness 0.93 1.00 – – – Variances Close relationships 0.91 1.37 0.04 39.70 <.001 Community belonging 0.84 0.72 0.03 27.55 <.001 Family trust 0.81 0.59 0.04 14.40 <.001 General health 0.59 0.67 0.03 23.08 <.001 Mental health 0.58 0.66 0.03 20.10 <.001 Life satisfaction 0.95 8.59 0.19 44.88 <.001 Neighbour trust 0.89 1.02 0.03 33.09 <.001 Neighbour connection 0.86 0.72 0.02 41.74 <.001 Neighbour helpfulness 0.14 0.02 0.01 2.99 <.001 Social support 1.00 0.14 0.02 7.38 <.001 Resilience 1.00 0.47 0.03 16.68 <.001 Community Cohesion 1.00 0.12 0.01 15.41 <.001 Note. Std. est = standardized estimate 𝛽. Est = unstandardized estimate B. Standard errors (SE) = standard errors of each estimate. 56 Table 26 Model 4R: R2 Values for Each Variable Manifest variable R2 Close relationships .09 Community belonging .16 Family trust .19 General health .42 Mental health .42 Life satisfaction .05 Neighbour trust .11 Neighbour connection .15 Neighbour helpfulness .86 Model 4B Relationships between Social Support, Community Cohesion, and Resilience with Freely Estimated Factor Loadings Another model was examined to determine the correlation between social support and community cohesion, as well as the direct paths between social support and resilience and community cohesion and resilience. For this model, the factor loadings were freely estimated. This model allowed for the possibility that each manifest variable had unequal relationships with their respective latent variable. 57 Figure 11 Model 4B: A Structural Regression Model of Community Cohesion, Resilience, and Social Support -Freely Estimated Factor Loadings Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols. 11Community CohesionD1 1NeighbourTrustNeighbourHelpfulnessNeighbourConnectionE11E2 E31 1Close RelationshipsCommunity BelongingFamily TrustD21Social SupportE41E5 E61 1ResilienceMental Health General Health Life SatisfactionE71E8 E31 1D311 58 Model 4B Results. There were 45 non-redundant observations – 21 parameter estimates = 24; dfm ≥ 0. The additional requirements for identification of SR models were also met. As such, Model 4B was overidentified. Recall that for the latent factors their variances were set to 1 with means set to 0. The parameter estimates were calculated using WLSMV estimation due to the dichotomous nature of the neighbour connection variable. As WLSMV estimator adjusts for non-normality, the robust parameter values are reported. While running this model, an error was identified. The correlation between social support and community cohesion was greater than 1. Given that model results from a model with a correlation between latent variables that is greater than 1 is unreliable, the model fit and parameters were not reported. Error Diagnostics. Due to the error that was found, analyses were conducted to determine the source of the error. First, this error may have been suggestive of multicollinearity. However, based on the correlations, VIF, and tolerance statistics that were reported above, multicollinearity among the data was unlikely. In addition, the eigen values were also calculated which indicated the strength of the relationship between a given factor and the set of observed variables. Eigenvalues close to 0 suggest multicollinearity (Pituch & Stevens, 2015). As such, based on the Eigenvalues, there was no evidence of multicollinearity within the data. When factors have eigenvalues well below 1, according to Kaiser’s rule, it would suggest that the manifest variables are not good indicators of the latent variables (Pituch & Stevens, 2015). Here, it is suggested that close relationships, community belonging, and family trust are all good measures of social support. However, life satisfaction may not be a good measure of resilience. In addition, neighbour trust, neighbour connection, and neighbour helpfulness, respectively, may not be good indicators of community cohesion. See Table 27 for eigenvalues. 59 Table 27 Model 4 and 4b: Social Support, Community Cohesion and Resilience Eigenvalues CR CB FT MH GH LS NT NC NH Eigenvalue 2.84 1.44 1.06 1.00 0.84 0.58 0.46 0.43 0.37 Note. CR = close relationships, CB = community belonging, MH = mental health, GH = general health, LS = life satisfaction, NT = neigbour trust, NC = neighbour connection, NH = neighbour helpfulness. CHAPTER 4 GENERAL DISCUSSION AND CONCLUSIONS The current study examined the relationships between social support, community cohesion, and resilience in adults who were exposed to physical P-IPV during childhood. Structural equation modelling was used to evaluate whether there were significant positive relationships between social support, community cohesion, and resilience, respectively. For the purposes of this study, resilience was operationalized using three manifest variables: general health, mental health, and life satisfaction. In general, structural equation modelling techniques appeared to be a sufficient method of addressing my first and third research questions as evidenced by good fitting models. However, SEM was not able to adequately address my fourth research question due to the highly correlated latent constructs of social support and community cohesion. Therefore, findings from this study reflect results obtained from my interpretation of parameter estimates from SEM analyses, Spearman correlations between manifest variables, the correlations between the manifest variables and their respective latent variables (i.e., the standardized estimates of the covariances), squared multiple correlations, and the eigenvalues for the final model. 60 Model Specific Discussion Findings and Discussion for Research Question 1: Relationship Between Social Support and Resilience The results from the SEM models for the first research question (i.e., Model 1R and Model 1B) suggest that the model fit the data well when the factor loadings were freely estimated. Overall, there was evidence from the current study to suggest that there is a positive direct relationship between social support (measured as close relationships, community belonging, and family trust) and resilience (measured as general health, mental health, and life satisfaction) for P-IPV exposed individuals. In addition, there is evidence to suggest that there is a moderate positive correlation between social support and resilience. This finding aligns with past research studies that have suggested a significant positive relationship between social support and resilience for this population (Osofsky, 1999; Yule et al., 2019). The methodological difference among the models of having fixed versus freely estimated factor loadings has important implications for the conceptualization of social support that should be considered in future research. Allowing the parameters to be freely estimated led to improved model fit compared to having fixed factor loadings. This suggests that the manifest variables of close relationships, community belonging, and family trust are not equal measures of social support (i.e., some of the manifest variables are better predictors of social support than others). Based on the Spearman’s correlations, there were significant, albeit weak, correlations among the indicator variables (close relationships, community belonging, and family trust) for the latent construct of social support. These correlations may have been biased as the significance level of correlations is highly influenced by large sample size (LaMorte, 2021). This suggests that although the manifest variables have significant impacts on one another, there are likely other 61 important determinants as well (LaMorte, 2021). Similarly, the squared multiple correlations suggested that the close relationships and community belonging variables each accounted for less than 20 percent of the variance in social support. As such, they were not adequate measures of social support and should have been considered for removal from the model (Hooper et al., 2008). Given the weak correlations between the manifest variables for social support, it can be conceived that the manifest variables of close relationships, community belonging, and family trust may not be the best indicators of social support. Importantly, there were conceptual differences in the measures of social support used in Osofsky (1999), Yule and colleagues (2019), and the current study. In the review conducted by Osofsky (1999), social support was measured as the presence of a loving and supportive relationship with a trusted adult (i.e., a neighbour, teacher, coach, parent, or grandparent). In addition, Osofsky (1999) examined exposure to violence by including exposure to chronic community violence, parental physical violence, and exposure to P-IPV. Similarly, the social support measures used by Yule et al. (2019) were family and peer support. I had initially proposed social support as being comprised of variables measuring family IPV disclosure, friend IPV disclosure, neighbour IPV disclosure, police IPV disclosure, and CAS IPV disclosure as proxies for informal and formal social support received. However, given that disclosure to family, friends, neighbours, and police was only measured for those who reported child abuse and/or neglect but not exposure to P-IPV, social support was reconceptualized. A limitation of this study was the limited selection of measured variables that could be used to create latent constructs. Given that a minimum of three manifest variables are needed to create a latent construct (Kline, 2015), and the limited variables available in the dataset, I was not able to 62 include manifest variables that perfectly aligned with the conceptualization of social support used in previous research. Relatedly, Coker et al. (2003) used SEM to analyze the promotive role of social support in resilience for survivors of IPV. Despite the difference in population (i.e., IPV survivors versus P-IPV exposed adults), Coker et al. (2003) found that social support significantly mediated the relationship between IPV and better mental and physical health. The measure of social support they used was the Social Support Questionnaire – Short Form (Sarason et al., 1987), which measured whether women felt like they had someone to count on when they needed them. Although the social support latent variable in the current study included a variable assessing the number of close relationships participants’ reported, this variable did not seem to represent social support well. This may be because only one item measured the number of participants’ close relationships in the current study compared to use of the multi-item Social Support Questionnaire – Short form in the Coker et al. (2003) study. Future research should seek to conceptualize social support using measures of close relationships that may include friend support, family support, neighbour support, and support of a trusted adult. Using a validated questionnaire to capture social support like the Social Support Questionnaire – Short Form (Sarason et al., 1987) may provide a more comprehensive measure of social support with increased standardization to compare results across studies. In addition, given that social support changes over the lifetime, it is likely that the social supports that were available to participants at the time of the P-IPV exposure (i.e., when they were children) were different from the social supports available to them as adults. For studies that wish to use adult samples with retrospective reporting of P-IPV exposure, it is recommended to measure both the social supports available to respondents at the time of the IPV exposure as well as present day 63 social supports. As there may be differential impacts of social support depending on the timing in which the support was received, the relationships between resilience and social support available at the time of P-IPV exposure versus during adulthood should be examined. Findings and Discussion Research Question 3: Relationship Between Community Cohesion and Resilience The results from the SEM models for the first research question (i.e., Model 1R and Model 1B) suggested that the model fit the data well when the factor loadings were freely estimated. Overall, there was evidence from the current study to suggest that there is a weak positive direct relationship between community cohesion (measured as neighbour trust, neighbour helpfulness, and neighbour connection) and resilience (measured as general health, mental health, and life satisfaction) for P-IPV exposed individuals. In addition, there was a weak positive correlation between community cohesion and resilience. These findings are in line with past research studies that have suggested a significant positive relationship between community cohesion and resilience (Fagan et al., 2014; Greenfield & Marks, 2010; Yule et al., 2019). There are important differences in study methodology between the current study and studies conducted by Yule et al. (2019), Fagan et al. (2014), and Greenfield and Marks (2010) that are worth noting. First, the sample used within each study varied. For the Yule and colleagues (2019) study, they included participants who had been exposed to violence and neglect that extended beyond P-IPV only. That is, participants could have been exposed to P-IPV, community violence, physical or sexual child maltreatment, and/or neglect. Similarly, Fagan and colleagues (2014) also used a sample of adolescents who had been exposed to community violence rather than P-IPV. As such, the current study extends these findings by providing 64 support for the relationship between community cohesion and resilience for individuals who were exposed exclusively to P-IPV. Second, Yule and colleagues (2019) separated studies in their review based on cross-sectional and longitudinal studies. For the longitudinal studies, they did not find a significant relationship between community cohesion and resilience for P-IPV exposed adults. Yule and colleagues (2019) suggested the difference in these findings (i.e., the significant relationship between community cohesion and resilience in the cross-sectional but not the longitudinal studies) was due to the large variability among study effect sizes. This difference may also be explained by inconsistent reporting that may happen over the course of time due in part to the adjustment process to violence that a person undergoes (McKinney et al., 2009). From child sexual abuse research, it has been found that people who have previously reported abuse may no longer report abuse years later as they may wish to put the abuse “behind them.” Additionally, in IPV research, it has been suggested that symptoms of depression and anxiety that may be experienced by survivors or children exposed to P-IPV may influence inconsistent reporting of IPV over time (Trevillion et al., 2012; Vezina & Hébert, 2007). Similar to the longitudinal studies in Yule et al. (2019), for the GSS data used in the current study, there were at least four years between the P-IPV exposure and the data collection (and likely many more years for most participants) as the P-IPV exposure was reported before age 15 and the youngest respondents were 18 years of age. Despite the limitations of using retrospective data, the relationship between community cohesion and resilience was significant. This further highlights the strength of the relationship between community cohesion and resilience. Greenfield and Marks (2010) also used retrospective reporting of P-IPV exposure. However, in their study, P-IPV exposure was measured using a series of items from a modified 65 version of the Conflict Tactics Scale (CTS; Straus, 1979) that measured physical and psychological P-IPV exposure. A questionnaire with multiple items is more likely to accurately capture P-IPV exposure than a single-item measure. As previously mentioned, the inclusion of respondents who have been exposed to P-IPV based on only one survey item is a limitation of the current study. In addition, Fagan and colleagues (2014) used the outcome measures of substance use and violence perpetration instead of resilience (i.e., mental health, general health, and life satisfaction). The current study helps to demonstrate the protective role of community cohesion after P-IPV exposure on general health, mental health, and life satisfaction (i.e., resilience). A strength of the current study was the conceptualization of community cohesion. Community cohesion has been previously defined as the presence of trustworthy, helpful, and involved neighbours (Herrero & Gracia, 2007). In the Fagan et al. (2014) study, they measured collective efficacy in which communities with high levels of collective efficacy are more likely to have adults and youth who know and trust each other and are more likely to help one another in an attempt to reduce crime and delinquent behaviours (Sampson et al., 1997). Similarly, Yule and colleagues (2019) included studies with the construct of community cohesion, which included collective efficacy, sense of security, and neighbour helpfulness, involvement, and trust. Riina (2021) also used a similar measure of social cohesion of communities in their study by using the validated Collective Efficacy Scale (CES) by Sampson et al. (1997). Social cohesion of communities was defined as having people in one’s communities who are willing to help their neighbours, having trust in their neighbours, having a tight knit neighbourhood, and neighbours getting along well with each other and having the same values (Riina, 2021). The construct of community cohesion in the current study paralleled this concept of collective efficacy by 66 measuring neighbour connection, neighbour helpfulness, and neighbour trust. Given the strong conceptualization of community cohesion that many researchers have used and the existence of the validated Collective Efficacy Scale (CES) by Sampson et al. (1997) to measure community cohesion, it is recommended that future research continue to incorporate these scales to capture this construct. It is proposed that future research focus on conducting longitudinal studies. Given the limitations of retrospective reporting, more longitudinal research is required to determine the impact of community cohesion on resilience in P-IPV exposed children prospectively. Similar to the sources of social support, it can be deduced that as individuals grow and move out of their familial home, the cohesiveness of their communities may differ. This can be done in a similar format to the GSS victimization studies in which surveys are completed by Canadian youth beginning at age 12 using developmentally appropriate language. These surveys can be administered every two years to measure differences in P-IPV exposure that may happen over time. In order to improve the method of identifying P-IPV exposed individuals, the Conflict Tactics Scale (CTS; Straus, 1979) or similar measures should be embedded in this survey. This would allow for the examination of reporting patterns of P-IPV over time by following cohorts throughout development and to examine the impact of community cohesion during the P-IPV exposure as well as post-P-IPV exposure. An added area of exploration may be to examine if P-IPV exposure during childhood as well as the presence of a cohesive community during childhood has an impact on which neighbourhoods individuals choose to settle down in (i.e., does P-IPV exposure make individuals more or less likely to seek out cohesive communities as an adult?). 67 Findings and Discussion Research Question 4: Relationships Among Social Support, Community Cohesion, and Resilience The results from the SEM for the fourth model suggested that the model did not fit the data well. Given the poor fit, it was difficult to answer the fourth research question. Overall, there was not strong evidence to suggest that there was a unique positive relationship between social support (measured as close relationships, community belonging, and family trust) and resilience (measured as general health, mental health, and life satisfaction); or community cohesion (measured as neighbour trust, neighbour connection, and neighbour helpfulness) and resilience, when accounting for the correlation between social support and community cohesion for P-IPV exposed individuals. In essence, the model did not suggest a collective influence of social support and community cohesion on resilience. In addition, the model cannot be used to determine whether social support or community cohesion is a stronger predictor of resilience. These findings are in opposition with past research studies that have suggested a collective influence of social support and community cohesion on resilience. Past research has demonstrated that community cohesion acts as a source of social support for children and their families (Sampson et al., 2002). One proposed impact of this social support is a decrease in IPV perpetration. Showalter and colleagues (2017) demonstrated that communities that were higher in informal social control (that accompanies collective efficacy) were associated with lower levels of experienced IPV among non-single mothers. Similarly, Caetano et al. (2010) examined a nationally representative sample of couples and found that self-reported social cohesion in neighbourhoods acted to significantly reduce male perpetrated IPV. However, despite these promising results of a decrease in IPV perpetration, and accompanying 68 P-IPV exposure, the current study did not find a significant relationship between social support and community cohesion. Based on the results, it is possible that the manifest variables used to measure social support and those to measure community cohesion were too highly correlated. In essence, it can be argued that some of the manifest variables making up the social support and community cohesion latent variables in the current models may be indicators of both social support and community cohesion. For instance, one manifest variable that was used to measure social support was community belonging. Based on the Spearman’s correlations, the correlation between neighbour connection and the manifest variables of community cohesion (i.e., neighbour trust, neighbour connection and neighbour helpfulness) were all stronger than the correlations between neighbour connection and the other social support variables (i.e., close relationships and family trust). It can be argued that this manifest variable is a stronger measure of community cohesion than social support. Relating this back to SAMSHA’s socio-ecological model (SAMSHA, 2014), it can be argued that the variable of community belonging fits better with the community and organizational factors level rather than the interpersonal factors level of their model. Based on the results from the current study combined with the evidence that community belonging aligns with the conceptualization of community cohesion put forth by Herrero and Gracia (2007), it is recommended that future research include community belonging as an indicator of community cohesion. In addition, in the literature more broadly, there is currently an absence of studies that demonstrates the indirect effect of community cohesion on resilience through social support. This is a phenomenon that requires future examination. Future research may wish to use validated measures like the Social Support Questionnaire – Short Form (Sarason et al., 1987) to measure social support and the Collective Efficacy Scale (CES) by Sampson et al. 69 (1997). The use of validated measures to examine the relationship between social support and community cohesion in P-IPV exposed children may help decrease the measurement overlap in the related constructs of social support and community cohesion. The Comparison of Relationships Between Social Support and Resilience and Community Cohesion and Resilience Overall, in the current study, the relationship between social support and resilience was found to be a stronger relationship than that of community cohesion and resilience. One possible reason for this is that the measure of social support included the indicator of community belonging. As it may be argued that community belonging fits better with the community and organizational level rather than the interpersonal factors level of the SAMSHA socio-ecological model (2014), the social support variable may have captured two levels of the SAMSHA model whereas community cohesion only captures one level. Another possible reason for this difference in strength between the relationships is that social support represents a proximal protective factor whereas community cohesion is a distal protective factor. According to Oatley and colleagues (2006) proximal protective factors are those that have a direct impact on the child and their functioning (e.g., parent-child relationships, attachment status, differential parenting among siblings) whereas distal risk factors are those that are more indirect that impact children’s functioning (e.g., neighbourhood support, low socioeconomic status, income inequality). It can be extrapolated that proximal protective factors (e.g., social support) would have a stronger impact on resilience than distal protective factors (e.g., community cohesion). Finally, given the strong connection between social support and community cohesion that has been demonstrated in previous research (e.g., Sampson et al., 2002), it is possible that community cohesion is best understood as one specific form of social support. It follows that a higher number of sources of 70 support (i.e., social support more broadly) would have an increased effect of buffering the negative impacts of P-IPV exposure and have a stronger impact on resilience than community cohesion alone. Future research is required to determine if community cohesion is best understood as a form of social support or as a distinct concept. Overall Limitations Capturing P-IPV Exposure First, the measure of P-IPV exposure may have been limited in its ability to capture those who had been exposed to P-IPV as children given that it consisted of only one question (i.e., “Before age 15, how many times did you see or hear any one of your parents, step-parents or guardians do any of the following? Hit each other).” For instance, of those who reported exposure to physical fights between their parental figures, over 40% reported that they were exposed one to two times. Thus, some participants were not exposed to ongoing P-IPV but were instead exposed to a small number of potentially isolated incidents. In addition, this conceptualization of P-IPV exposure did not include psychological or emotional P-IPV exposure which may be equally or more insidious for child development (Fritz & Roy, 2022). Moreover, we may have excluded some participants who in fact had been exposed to P-IPV but did not report the exposure (given past research that suggests that underreporting of P-IPV is common; Public Health Agency of Canada, 2018). As such, there is the potential that this sample was not an accurate representation of adults who were exposed to P-IPV during childhood. Measuring Resilience In the current study, the latent construct of resilience was conceptualized using the measured indicators of general health, mental health, and life satisfaction. Based on the Spearman correlations, the correlations between general and mental health were moderate. Similarly, the squared multiple correlations suggest that general health accounted for roughly 71 40% of the variance in each model, whereas mental health accounted for roughly 40-70% of the variance in each model. Given that both mental health and general health accounted for over 20% of the variance in each model, they are considered good measures of resilience (Hooper et al., 2008). This aligns with past research that demonstrates strong positive correlations between mental health, physical health, and resilience, respectively (Nath & Pradhan, 2012). The correlation between life satisfaction and general health was moderate whereas the correlation between life satisfaction and mental health was weak. The squared multiple correlations for each model suggested that life satisfaction may not have accounted for any (0%) of the variance in the model. Taken together, these findings suggests that life satisfaction was not likely an adequate indicator of resilience. When compared to the conceptualizations of resilience used in previous studies, most commonly, researchers use the absence of internalizing and externalizing symptoms (e.g., Riina, 2021) or levels of psychopathology more broadly (Masten & Reed, 2002) as indicators of resilience. As has been suggested by past research, this study used the definitional elements to capture resilience (i.e., having experience an adversity—P-IPV exposure—and exhibiting positive functioning despite adversity [here mental health, general health, and life satisfaction]). This study aimed to broaden the definitional element of positive adaptation despite adversity by including life satisfaction as an indicator of positive adaptation (Fletcher & Sarkar, 2013; Masten et al., 1990). However, the current study did not show that life satisfaction was an adequate measure of positive adaptation. Although past research has demonstrated a significant positive relationship between life satisfaction and resilience (Tamarit et al., 2023; Zheng et al., 2020), it can be argued that life satisfaction and resilience are conceptually different. Life satisfaction (or perceived life satisfaction) has been previously defined as a person viewing themselves as valuable and as 72 believing that their life has meaning (Batmaz & Çelik, 2015). This valuation of one’s value and satisfaction with their life circumstances differs from adapting well despite exposure to challenge or adversity (i.e., resilience). Life satisfaction does not require the presence of a challenge or adversity in the same way that resilience does. As such, future research should explore the relation between life satisfaction and resilience. Findings from the current study suggest that it may be more beneficial to include a comprehensive and validated measure of resilience, such as the Connor Davidson Resilience Scale (CDRISC; Connor & Davidson, 2003), when representing resilience in SEM rather than measures of life satisfaction. Strengths The strengths of the present study included the use of a nationally representative dataset. The use of the GSS dataset allowed for a high level of external validity (i.e., generalizability) of results for Canadian adults who were exposed to P-IPV during childhood. Another strength was the large sample size, which allowed the data to be analyzed using SEM and for the use of statistical procedures to address issues related to missing data (e.g., FIML), non-normality (e.g., MLR), dichotomous indicator variables (e.g., WLSMV), etc. Finally, all models were proposed a priori based on prior research, and revisions to the models were made with prior theory and research in mind. In a similar vein, modification indices were not used to modify the models in an effort to stay true to the originally conceptualized models. Practical Importance of Continued Research on Resilience in P-IPV Exposed Individuals In 2018 in Canada alone, there were 17,051 substantiated investigations of childhood exposure to IPV (Fallon et al., 2015). The widespread impact of P-IPV exposure was evident in the current study as over 10 percent of 2019 GSS Victimization survey respondents reported physical P-IPV exposure. Indeed, over 600 respondents experienced chronic P-IPV exposure as 73 they reported experiencing more than 10 physical P-IPV exposures before the age of 15 years. Based on previous research of the negative sequalae of P-IPV exposure (e.g., Bogat et al., 2023; Fritz & Roy, 2022), P-IPV exposure is impacting the development and long-term outcomes of many Canadians. Research examining resilience among individuals exposed to P-IPV is in its infancy relative to other IPV research. Given the high external validity of the current study given the nationally representative sample of Canadian adults, findings suggest that future research in this area is warranted to help improve the health and wellbeing of many Canadians. Findings suggest that the social support and community cohesion that adult survivors of P-IPV have established during adulthood—even many years after their exposure to P-IPV—are positively associated with their resilience. It is thus important to develop an understanding of the impact of social support and community cohesion on resilience, including how to create stronger support networks and more cohesive communities, to negate some of the insidious effects of P-IPV exposure. An important implication of this research is a call to action to create resources that help individuals develop stronger social supports and more cohesive communities. Additionally, more distal recommendations include educating municipal governments on the importance of allocating funds to create safer, more cohesive communities by increasing events that foster community (e.g., local holiday events such as parades and celebrations for communities including opportunities for individuals to volunteer during these events). For those who have daily interactions with individuals who may have experienced P-IPV exposure, it is important that they encourage these individuals to become involved with social groups and community groups and to help foster those connections by sharing appropriate resources (e.g., psychologists sharing resources of where their clients may join local community groups that match their 74 interests). Policy makers may also be educated on the importance of the more distal benefits of funding social/community initiatives, which may include buffering the impact of violence on citizens as well as a decrease in violence (e.g., IPV) occurrence itself. The continued examination of resilience in P-IPV exposed individuals is imperative. An individual’s functioning and being at risk of experiencing the negative sequalae of P-IPV exposure are not mutually exclusive. Someone can appear to be functioning well (i.e., demonstrating resilience) and have an abundance of protective factors present in their lives but continue to be at risk if they are still witnessing parental IPV. As such, the top priority for working with IPV exposed individuals should be ongoing risk assessment and safety planning (Jaffe, 2014). Overall, future research on resilience in P-IPV exposed individuals may provide the foundational knowledge that is required to equip professionals who interact with P-IPV exposed individuals (e.g., teachers, social workers, relatives, psychologists, police officers) with the knowledge and skills to provide high quality support to P-IPV exposed individuals in ways that will foster resilience (e.g., underscoring the importance of social support networks, seeking ways for individuals to be more integrated into their communities). In addition, research suggests that there are shared risk factors for children exposed to violence, regardless of the type of violence experienced (e.g., child maltreatment, community or neighbourhood violence, and exposure to IPV; Hamby & Grych, 2013). As such, the findings may benefit individuals exposed to violence more generally, regardless of the type of violence they experienced (Tabibi et al., 2020). 75 References Alaggia, R., & Donohue, M. (2018). Take these broken wings and learn to fly: Applying resilience concepts to practice with children and youth exposed to intimate partner violence. Smith College Studies in Social Work, 88(1), 20–38. https://doi.org/10.1080/00377317.2018.1404282 Anderson, K. M., & Bang, E.-J. (2012). Assessing PTSD and resilience for females who during childhood were exposed to domestic violence. Child & Family Social Work, 17(1), 55–65. https://doi.org/10.1111/j.1365-2206.2011.00772.x Anderson, K. M., Renner, L. M., & Danis, F. S. (2012). Recovery: Resilience and growth in the aftermath of domestic violence. Violence against women, 18(11), 1279–1299. https://doi.org/10.1177/1077801212470543 Bair-Merritt, M. H., Blackstone, M., & Feudtner, C. (2006). Physical health outcomes of childhood exposure to intimate partner violence: A systematic review. Pediatrics, 117(2), e278-90. https://doi.org/10.1542/peds.2005-1473 Baker, L., Straatman, A.L., Etherington, N., O’Neil, B., Heron, C., Sapardanis, K. (2016). Towards a conceptual framework: Trauma, family violence and health. London, ON: Knowledge Hub, Learning Network, Centre for Research & Education on Violence against Women & Children. https://www.learningtoendabuse.ca/resources-events/pdfs/Framework_paper-April-20171.pdf Bartlett, J. & Steber, K. (2019) How to implement trauma-informed care to build resilience to childhood trauma - Child trends. [online] Child Trends. https://www.childtrends.org/publications/how-to-implement-trauma-informed-care-to-build-resilience-to-childhood-trauma 76 Batmaz, H., & Çelik, E. (2024). Examining the mediating role of resilience and life satisfaction in the relationship between anxiety sensitivity and perceived stress. Psychological Reports. Advance online publication. h%ps://doi.org/10.1177/00332941241263572 Belsky, J., Bakermans-Kranenburg, M. J., & van IJzendoorn, M. H. (2007). For better and for worse. Current Directions in Psychological Science, 16(6), 300–304. https://doi.org/10.1111/j.1467-8721.2007.00525.x Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107(2), 238–246. https://doi.org/10.1037/0033-2909.107.2.238 Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 88(3), 588–606. https://doi.org/10.1037/0033-2909.88.3.588 Black, C., & Ford-Gilboe, M. (2004). Adolescent mothers: Resilience, family health work and health-promoting practices. Journal of Advanced Nursing, 48(4), 351–360. https://doi.org/10.1111/j.1365-2648.2004.03204.x Black, T., Fallon, B., Nikolova, K., Tarshis, S., Baird, S., & Carradine, J. (2020). Exploring subtypes of children’s exposure to intimate partner violence. Children and Youth Services Review, 118(105375), 105375. https://doi.org/10.1016/j.childyouth.2020.105375 Bogat, G. A., Levendosky, A. A., & Cochran, K. (2023). Developmental consequences of intimate partner violence on children. Annual Review of Clinical Psychology, 19, 303–329. https://doi.org/10.1146/annurev-clinpsy-072720-013634 Boyce, W. T., & Ellis, B. J. (2005). Biological sensitivity to context: I. An evolutionary-developmental theory of the origins and functions of stress reactivity. Development and Psychopathology, 17(2), 271–301. https://doi.org/10.1017/s0954579405050145 77 Bozdogan, H. (1987). Model selection and Akaike’s information criterion (AIC): The general theory and its analytical extensions. Psychometrika, 52(3), 345–370. Browne, M. W., & Cudeck, R. (1989). Single sample cross-validation indices for covariance structures. Multivariate Behavioral Research, 24(4), 445–455. https://doi.org/10.1207/s15327906mbr2404_4 Butchart, A., World Health Organization(WHO), & Mikton, C. (2014). Global status report on violence prevention. World Health Organization. Caetano, R., Ramisetty-Mikler, S., & Harris, T. R. (2010). Neighborhood characteristics as predictors of male to female and female to male partner violence. Journal of Interpersonal Violence, 25(11), 1986–2009. https://doi.org/10.1177/0886260509354497 Calhoun, L. G., & Tedeschi, R. G. (2014). The foundations of posttraumatic growth: An expanded framework. In Calhoun, L. G., & Tedeschi, R. G. Handbook of posttraumatic growth (pp. 3-23). Routledge. Chan, K. L., Chen, M., Chen, Q., & Ip, P. (2017). Can family structure and social support reduce the impact of child victimization on health-related quality of life? Child Abuse & Neglect, 72, 66–74. https://doi.org/10.1016/j.chiabu.2017.07.014 Chan, Y.-C., & Yeung, J. W.-K. (2009). Children living with violence within the family and its sequel: A meta-analysis from 1995–2006. Aggression and Violent Behavior, 14(5), 313–322. https://doi.org/10.1016/j.avb.2009.04.001 Coker, A. L., Watkins, K. W., Smith, P. H., & Brandt, H. M. (2003). Social support reduces the impact of partner violence on health: Application of structural equation models. Preventive Medicine, 37(3), 259-267. https://doi.org/10.1016/S0091-7435(03)00122-1 78 Connor, K. M., & Davidson, J. R. (2003). Development of a new resilience scale: The Connor‐Davidson Resilience Scale (CD‐RISC). Depression and Anxiety, 18(2), 76-82. https://doi.org/10.1002/da.10113 Conroy, S. (2019). Section 3: Police-reported intimate partner violence in Canada, 2019. Family violence in Canada: A statistical profile, 29-42. Cotter, A. (2021). Intimate partner violence in Canada, 2018: An overview. Juristat. Statistics Canada Catalogue no. 85-002-X. https://www150.statcan.gc.ca/n1/pub/85-002-x/2021001/article/00003-eng.htm Cunningham, A. & Baker, L. (2007). Little eyes, little ears: How violence against a mother shapes children as they grow. London, ON: Centre for Children & Families in the Justice System. https://www.canada.ca/en/public-health/services/health-promotion/stop-family-violence/prevention-resource-centre/women/little-eyes-little-ears-violence-against-a-mother-shapes-children-they-grow.html Davvetas, V., Diamantopoulos, A., Zaefarian, G., & Sichtmann, C. (2020). Ten basic questions about structural equations modeling you should know the answers to–But perhaps you don't. Industrial Marketing Management, 90, 252-263. https://doi.org/10.1016/j.indmarman.2020.07.016 Ehrensaft, M. K., Knous-Westfall, H., & Cohen, P. (2017). Long-term influence of intimate partner violence and parenting practices on offspring trauma symptoms. Psychology of Violence, 7(2), 296–305. https://doi.org/10.1037/a0040168 Ellis, B. J., & Boyce, W. T. (2011). Differential susceptibility to the environment: Toward an understanding of sensitivity to developmental experiences and context. Development and Psychopathology, 23(1), 1–5. https://doi.org/10.1017/S095457941000060X 79 Enders, C. K. (2022). Applied missing data analysis. Guilford Publications. English, D. J., Marshall, D. B., & Stewart, A. J. (2003). Effects of family violence on child behavior and health during early childhood. Journal of Family Violence, 18(1), 43-57. Ernst, A. A., Weiss, S. J., Del Castillo, C., Aagaard, J., Marvez-Valls, E., D’Angelo, J., Combs, S., Feuchter, A., Hegyi, M., Clark, R., & Coffman, B. (2007). Witnessing intimate partner violence as a child does not increase the likelihood of becoming an adult intimate partner violence victim. Academic Emergency Medicine: Official Journal of the Society for Academic Emergency Medicine, 14(5), 411–418. https://doi.org/10.1197/j.aem.2006.11.032 Evans, S. E., Davies, C., & DiLillo, D. (2008). Exposure to domestic violence: A meta-analysis of child and adolescent outcomes. Aggression and Violent Behavior, 13(2), 131–140. https://doi.org/10.1016/j.avb.2008.02.005 Fagan, A. A., Wright, E. M., & Pinchevsky, G. M. (2014). The protective effects of neighborhood collective efficacy on adolescent substance use and violence following exposure to violence. Journal of Youth and Adolescence, 43, 1498-1512. https://doi.org/10.1007/s10964-013-0049-8 Fallon, B., Van Wert, M., Trocmé, N., MacLaurin, B., Sinha, V., Lefebvre, R., & Rha, W. (2015). Ontario incidence study of reported child abuse and neglect 2013. Toronto, ON: Canadian Child Welfare Research Portal. Fletcher, D., & Sarkar, M. (2013). Psychological resilience. European Psychologist, 18(1), 12–23. https://doi.org/10.1027/1016-9040/a000124 80 Fong, V. C., Hawes, D., & Allen, J. L. (2019). A systematic review of risk and protective factors for externalizing problems in children exposed to intimate partner violence. Trauma, Violence & Abuse, 20(2), 149–167. https://doi.org/10.1177/1524838017692383 Fritz, P. A. T., & Roy, E. (2022). Relation between exposure to parental intimate partner violence during childhood and children’s functioning. In Geffner, R., White, J.W., Hamberger, K.L., Rosenbaum, A., Vaughan-Eden, V., & Vieth, V. Handbook of Interpersonal Violence and Abuse Across the Lifespan (pp. 2463–2483). Springer International Publishing. https://doi.org/10.1007/978-3-319-89999-2_312 Galano, M. M., Grogan-Kaylor, A., Clark, H. M., Stein, S. F., & Graham-Bermann, S. A. (2021). Examining the 8-year trajectory of posttraumatic stress symptoms in children exposed to intimate partner violence. Journal of Interpersonal Violence, 36(15–16), NP8454–NP8481. https://doi.org/10.1177/0886260519844280 Gardner, M. J., Thomas, H. J., & Erskine, H. E. (2019). The association between five forms of child maltreatment and depressive and anxiety disorders: A systematic review and meta-analysis. Child Abuse & Neglect, 96(104082), 104082. https://doi.org/10.1016/j.chiabu.2019.104082 Gewirtz, A. H., & Edleson, J. L. (2007). Young children’s exposure to intimate partner violence: Towards a developmental risk and resilience framework for research and intervention. Journal of Family Violence, 22(3), 151–163. https://doi.org/10.1007/s10896-007-9065-3 Government of Canada (2022). 2019 General Social Survey – Canadians’ safety: Technical report. Government of Canada, Statistics Canada. https://www.statcan.gc.ca/en/statistical-programs/document/4504_D1_V1 81 Greenfield, E. A., & Marks, N. F. (2010). Sense of community as a protective factor against long-term psychological effects of childhood violence. The Social Service Review, 84(1), 129–147. https://doi.org/10.1086/652786 Grych, J. H., & Fincham, F. D. (1990). Marital conflict and children’s adjustment: A cognitive-contextual framework. Psychological Bulletin, 108(2), 267–290. https://doi.org/10.1037/0033-2909.108.2.267 Hamby, S. L., & Grych, J. H. (2013). The web of violence: Exploring connections among different forms of interpersonal violence and abuse. Springer. Herrero, J., & Gracia, E. (2007). Measuring perceived community support: Factorial structure, longitudinal invariance, and predictive validity of the PCSQ (Perceived Community Support Questionnaire). Journal of Community Psychology, 35(2), 197–217. https://doi.org/10.1002/jcop.20143 Herrman, H., Stewart, D. E., Diaz-Granados, N., Berger, E. L., Jackson, B., & Yuen, T. (2011). What is resilience? Canadian Journal of Psychiatry. Revue Canadienne de Psychiatrie, 56(5), 258–265. https://doi.org/10.1177/070674371105600504 Holt, S., Buckley, H., & Whelan, S. (2008). The impact of exposure to domestic violence on children and young people: A review of the literature. Child Abuse & Neglect, 32(8), 797–810. https://doi.org/10.1016/j.chiabu.2008.02.004 Hooper, D., Coughlan, J., & Mullen, M. R. (2008). Structural equation modelling: Guidelines for determining model fit. The Electronic Journal of Business Research Methods, 6, 53-60. https://www.semanticscholar.org/paper/Structural-equation-modelling%3A-guidelines-for-model-Hooper-Coughlan/e55589eb030edbda5ce4035c1e4dc3cd1c7d1fdc 82 Howell, K. H. (2011). Resilience and psychopathology in children exposed to family violence. Aggression and Violent Behavior, 16(6), 562–569. https://doi.org/10.1016/j.avb.2011.09.001 Howell, K. H., Barnes, S. E., Miller, L. E., & Graham-Bermann, S. A. (2016). Developmental variations in the impact of intimate partner violence exposure during childhood. Journal of Injury & Violence Research, 8(1), 43–57. https://doi.org/10.5249/jivr.v8i1.663 Howell, K. H., & Miller-Graff, L. E. (2014). Protective factors associated with resilient functioning in young adulthood after childhood exposure to violence. Child Abuse & Neglect, 38(12), 1985–1994. https://doi.org/10.1016/j.chiabu.2014.10.010 Howell, K. H., Thurston, I. B., Schwartz, L. E., Jamison, L. E., & Hasselle, A. J. (2018). Protective factors associated with resilience in women exposed to intimate partner violence. Psychology of Violence, 8(4), 438–447. https://doi.org/10.1037/vio0000147 Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1-55. https://doi.org/10.1080/10705519909540118 Jackson, D. L. (2003). Revisiting sample size and number of parameter estimates: Some support for the N:Q hypothesis. Structural Equation Modeling: A Multidisciplinary Journal, 10(1), 128–141. https://doi.org/10.1207/s15328007sem1001_6 Jaffe, P. (2014). Risk factors for children in situations of family violence in the context of separation and divorce [electronic resource]. Department of Justice. https://www.justice.gc.ca/eng/rp-pr/cj-jp/fv-vf/rfcsfv-freevf/rfcsfv-freevf.pdf 83 Jaffe, P. G., Campbell, M., Hamilton, L. H. A., & Juodis, M. (2012). Children in danger of domestic homicide. Child Abuse & Neglect, 36(1), 71–74. https://doi.org/10.1016/j.chiabu.2011.06.008 Jenkins, J. (2008). Psychosocial adversity and resilience. In M. Rutter, D. Bishop, D. Pine, S. Scott, J. Stevenson, E. Taylor, & A. Thapar (Eds.), Rutter's child and adolescent psychiatry (pp. 377–391). Wiley Blackwell. https://doi.org/10.1002/9781444300895.ch25 Jöreskog, K. G., Sörbom, D. (1989). LISREL 7: User’s Reference Guide. Chicago, IL: Scientific Software. Kelleher, K. J., Hazen, A. L., Coben, J. H., Wang, Y., McGeehan, J., Kohl, P. L., & Gardner, W. P. (2008). Self-reported disciplinary practices among women in the child welfare system: Association with domestic violence victimization. Child Abuse & Neglect, 32(8), 811–818. https://doi.org/10.1016/j.chiabu.2007.12.004 Kimball, E. (2016). Edleson revisited: Reviewing children’s witnessing of domestic violence 15 years later. Journal of Family Violence, 31(5), 625–637. https://doi.org/10.1007/s10896-015-9786-7 Kitzmann, K. M., Gaylord, N. K., Holt, A. R., & Kenny, E. D. (2003). Child witnesses to domestic violence: A meta-analytic review. Journal of Consulting and Clinical Psychology, 71(2), 339–352. https://doi.org/10.1037/0022-006x.71.2.339 Kline, R. B. (2015). Principles and practice of structural equation modeling (4th ed.). Guilford Publications. Koenig, H. G. (2010). Spirituality and mental health. International Journal of Applied Psychoanalytic Studies, 7(2), 116-122. https://doi.org/10.1002/aps.239 84 Korkmaz, S., Goksuluk, D., & Zararsiz, G. (2014). MVN: An R package for assessing multivariate normality. The R Journal, 6(2), 151. https://doi.org/10.32614/rj-2014-031 Kralik, D., van Loon, A., & Visentin, K. (2006). Resilience in the chronic illness experience. Educational Action Research, 14(2), 187–201. https://doi.org/10.1080/09650790600718035 Laing, L., Humphreys, C., & Cavanagh, K. (2013). Social work and domestic violence developing critical & reflective practice. SAGE Publications. https://doi.org/ 10.4135/9781529681352 LaMorte, W. (2021). PH717 Module 9 - Correlation and Regression. Boston University School of Public Health. https://sphweb.bumc.bu.edu/otlt/MPH-Modules/PH717-QuantCore/PH717-Module9-Correlation-Regression/PH717-Module9-Correlation-Regression4.html Latzman, N. E., Vivolo-Kantor, A. M., Clinton-Sherrod, A. M., Casanueva, C., & Carr, C. (2017). Children’s exposure to intimate partner violence: A systematic review of measurement strategies. Aggression and Violent Behavior, 37, 220–235. https://doi.org/10.1016/j.avb.2017.10.009 Lee, I., Lee, E.-O., Kim, H. S., Park, Y. S., Song, M., & Park, Y. H. (2004). Concept development of family resilience: A study of Korean families with a chronically ill child. Journal of Clinical Nursing, 13(5), 636–645. https://doi.org/10.1111/j.1365-2702.2004.00845.x Levendosky, A. A., Bogat, G. A., Theran, S. A., Trotter, J. S., von Eye, A., & Davidson, W. S., 2nd. (2004). The social networks of women experiencing domestic violence. American 85 Journal of Community Psychology, 34(1–2), 95–109. https://doi.org/10.1023/b:ajcp.0000040149.58847.10 Little, R. J. A. (1988). Missing-data adjustments in large surveys. Journal of Business & Economic Statistics: A Publication of the American Statistical Association, 6(3), 287. https://doi.org/10.2307/1391878 Luthar, S. S., Cicchetti, D., & Becker, B. (2000). The construct of resilience: A critical evaluation and guidelines for future work. Child Development, 71(3), 543–562. https://doi.org/10.1111/1467-8624.00164 Mardia, K. V. (1970). Measures of multivariate skewness and kurtosis with applications. Biometrika, 57(3), 519-530. https://doi.org/10.1093/biomet/57.3.519 Margolin, G., & Gordis, E. B. (2000). The effects of family and community violence on children. Annual Review of Psychology, 51(1), 445–479. https://doi.org/10.1146/annurev.psych.51.1.445 Margolin, G., & Gordis, E. B. (2003). Co-occurrence between marital aggression and parents’ child abuse potential: The impact of cumulative stress. Violence and Victims, 18(3), 243–258. https://doi.org/10.1891/vivi.2003.18.3.243 Martinez-Torteya, C., Bogat, A., G., von Eye, A., & Levendosky, A. A. (2009). Resilience among children exposed to domestic violence: The role of risk and protective factors. Child Development, 80(2), 562–577. https://doi.org/10.1111/j.1467-8624.2009.01279.x Masten, A. S. (2001). Ordinary magic: Resilience processes in development. The American Psychologist, 56(3), 227–238. https://doi.org/10.1037/0003-066x.56.3.227 86 Masten, A. S. (2007). Resilience in developing systems: Progress and promise as the fourth wave rises. Development and Psychopathology, 19(3), 921–930. https://doi.org/10.1017/S0954579407000442 Masten, A. S. (2011). Resilience in children threatened by extreme adversity: Frameworks for research, practice, and translational synergy. Development and Psychopathology, 23(2), 493–506. https://doi.org/10.1017/S0954579411000198 Masten, A. S. (2015). Pathways to integrated resilience science. Psychological Inquiry, 26(2), 187–196. https://doi.org/10.1080/1047840x.2015.1012041 Masten, A.S. (2013). Risk and resilience in development. In P.D. Zelazo (Ed.), Oxford handbook of developmental psychology: Vol.2. Self and other (pp.579 - 607). Oxford University Press. https://doi.org/10.1111/jftr.12255 Masten, A. S., & Barnes, A. J. (2018). Resilience in children: Developmental perspectives. Children (Basel, Switzerland), 5(7). https://doi.org/10.3390/children5070098 Masten, A. S., Best, K. M., & Garmezy, N. (1990). Resilience and development: Contributions from the study of children who overcome adversity. Development and Psychopathology, 2(4), 425–444. https://doi.org/10.1017/s0954579400005812 Masten, A. S., & Cicchetti, D. (2016). Resilience in development: Progress and transformation. In D. Cicchetti (Ed.), Developmental psychopathology: Risk, resilience, and intervention (pp. 271–333). John Wiley & Sons, Inc. https://doi.org/10.1002/9781119125556.devpsy406 Masten, A. S., & Reed, M. J. (2002). Resilience in development. In R. Snyder & S. J. Lopez (Eds.), The Oxford handbook of positive psychology (pp. 74–88). Oxford University Press. 87 McIntosh, J. E., Tan, E. S., Levendosky, A. A., & Holtzworth-Munroe, A. (2021). Mothers’ experience of intimate partner violence and subsequent offspring attachment security ages 1–5 years: A meta-analysis. Trauma, Violence, & Abuse, 22(4), 885-899. https://doi.org/10.1177/1524838019888560 McKinney, C. M., Harris, T. R., & Caetano, R. (2009). Reliability of self-reported childhood physical abuse by adults and factors predictive of inconsistent reporting. Violence and Victims, 24(5), 653-658. https://doi.org/10.1891/0886-6708.24.5.653 McTavish, J. R., MacGregor, J. C. D., Wathen, C. N., & MacMillan, H. L. (2016). Children’s exposure to intimate partner violence: An overview. International Review of Psychiatry (Abingdon, England), 28(5), 504–518. https://doi.org/10.1080/09540261.2016.1205001 Muthén, B., & Muthén, L. K. (2000). Integrating person-centered and variable-centered analyses: Growth mixture modeling with latent trajectory classes. Alcoholism, Clinical and Experimental Research, 24(6), 882–891. https://doi.org/10.1097/00000374-200006000-00020 Nath, P., & Pradhan, R. K. (2012). Influence of positive affect on physical health and psychological well-being: Examining the mediating role of psychological resilience. Journal of Health Management, 14(2), 161-174. https://doi.org/10.1177/09720634120140020 Newsom, J. (2023) SEM with Categorical Variables. Psy 523/623 Structural Equation Modeling, Spring 2023. https://web.pdx.edu/~newsomj/semclass/ho_categorical.pdf Nikolova, K., Fallon, B., Black, T., Passanha, N., & Isaac, K. (2021). Responding to intimate partner violence (IPV) in Ontario, Canada: A closer look at police involvement. Children 88 and Youth Services Review, 128, Article 106168. https://doi.org/10.1016/j.childyouth.2021.106168 Noonan, C. B., & Pilkington, P. D. (2020). Intimate partner violence and child attachment: A systematic review and meta-analysis. Child Abuse & Neglect, 109(104765), 104765. https://doi.org/10.1016/j.chiabu.2020.104765 Nouhi, S., Janbozorgi, M., Heris, M. A., Najimi, A., & Khastar, H. (2017). Effectiveness of multi-faceted training focused on spiritual parenting to mothers on promoting spirituality of overweight children. International Journal of Health Studies, 3(1), 40-46. https://doi.org/10.22100/ijhs.v3i1.204 Oatley, K., Keltner, D., & Jenkins, J. M. (2006). Understanding emotions. Blackwell publishing. Onyskiw, J. E. (2003). Domestic violence and children’s adjustment. Journal of Emotional Abuse, 3(1–2), 11–45. https://doi.org/10.1300/j135v03n01_02 Osofsky, J. D. (1999). The impact of violence on children. The Future of Children, 9(3), 33–49. https://doi.org/10.2307/1602780 Owen, A. E., Thompson, M. P., Mitchell, M. D., Kennebrew, S. Y., Paranjape, A., Reddick, T. L., Hargrove, G. L., & Kaslow, N. J. (2008). Perceived social support as a mediator of the link between intimate partner conflict and child adjustment. Journal of Family Violence, 23(4), 221–230. https://doi.org/10.1007/s10896-007-9145-4 Perry, B. D. (1997). Incubated in terror: Neurodevelopmental factors in the “cycle of violence”. In J. D. Osofsky (Ed.), Children in a violent society (pp. 124-149). The Guilford Press. Pituch, K. A., & Stevens, J. P. (2015). Applied multivariate statistics for the social sciences (6th ed.). Routledge. 89 Public Health Agency of Canada. (2018, May 31). Family violence: How big is the problem in Canada? https://www.canada.ca/en/public-health/services/health-promotion/stop-family-violence/problem-canada.html Riina, E. M. (2021). Intimate partner violence and child and adolescent adjustment: The protective roles of neighborhood social processes. Journal of Family Psychology, 35(6), 756-766. https://doi.org/10.1037/fam0000893 Rosseel, Y. (2012). lavaan: An R package for structural equation modeling. Journal of Statistical Software, 48(2), 1–36. https://doi.org/10.18637/jss.v048.i02. Rubin, D. B. (1976). Inference and missing data. Biometrika, 63(3), 581-592. https://doi.org/10.2307/2335739 Sampson, R. J., Morenoff, J. D., & Gannon-Rowley, T. (2002). Assessing “neighborhood effects”: Social processes and new directions in research. Annual Review of Sociology, 28(1), 443–478. https://doi.org/10.1146/annurev.soc.28.110601.141114 Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277(5328), 918-924. Https://doi.org/10.1126/science.277.5328.918 Sarason, I. G., Sarason, B. R., Shearin, E. N., & Pierce, G. R. (1987). A brief measure of social support: Practical and theoretical implications. Journal of Social and Personal Relationships, 4(4), 497-510. https://doi.org/10.1177/0265407587044007 SAS Institute Inc. (2019). The MIANALYZE Procedure. SAS help center. https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_mianalyze_examples13.htm 90 Satyen, L., Rogic, A. C., & Supol, M. (2019). Intimate partner violence and help-seeking behaviour: A systematic review of cross-cultural differences. Journal of Immigrant and Minority Health, 21(4), 879–892. https://doi.org/10.1007/s10903-018-0803-9 Scott, K., Baker, L., Jenney, A., Lopez, J., Straatman, A.L., Antwi-Mansah, D., Cullen, O., Jones, K., Pietsch, N., and Expert Working Group Members. (2022). Recognizing critical expertise: A knowledge and skills framework for intimate partner violence specialists. London, ON: Centre for Research & Education on Violence Against Women and Children. https://www.learningtoendabuse.ca/docs/GBVExpertiseReport-June2022.pdf Serfilippi, E., & Ramnath, G. (2018). Resilience measurement and conceptual frameworks: A review of the literature. Annals of Public and Cooperative Economics, 89(4), 645–664. https://doi.org/10.1111/apce.12202 Showalter, K., Maguire-Jack, K., & Barnhart, S. (2017). Investigating the potentially protective effects of neighborhood processes in intimate partner violence. Journal of Aggression, Maltreatment & Trauma, 26(10), 1090-1103. https://doi.org/10.1080/10926771.2017.1320345 Sinha, M. (2013). Measuring violence against women: Statistical trends. Juristat: Canadian Centre for Justice Statistics, 1. Statistics Canada. (2015). Family Violence in Canada: A statistical profile, 2015. (Catalogue number 85-002-X). https://www150.statcan.gc.ca/n1/en/pub/85-002-x/2017001/article/14698-eng.pdf?st=Iv8M76YS Statistics Canada. (2021). Cycle 34: Canadians' safety (victimization) microdata user guide. Statistics Canada. 91 Straus, M. A. (1979). Conflict Tactics Scales (CTS) [Database record]. APA PsycTests. https://doi.org/10.1037/t02125-000 Stein, M. B., Choi, K. W., Jain, S., Campbell-Sills, L., Chen, C.-Y., Gelernter, J., He, F., Heeringa, S. G., Maihofer, A. X., Nievergelt, C., Nock, M. K., Ripke, S., Sun, X., Kessler, R. C., Smoller, J. W., & Ursano, R. J. (2019). Genome-wide analyses of psychological resilience in U.S. Army soldiers. American Journal of Medical Genetics. Part B, Neuropsychiatric Genetics: The Official Publication of the International Society of Psychiatric Genetics, 180(5), 310–319. https://doi.org/10.1002/ajmg.b.32730 Sternberg, K., Baradaran, L., Abbott, C., Lamb, M., & Guterman, E. (2006). Type of violence, age, and gender differences in the effects of family violence on children’s behavior problems: A mega-analysis. Developmental Review: DR, 26(1), 89–112. https://doi.org/10.1016/j.dr.2005.12.001 Substance Abuse and Mental Health Services Administration. (2014). SAMHSA’s concept of trauma and guidance for a trauma-informed approach. HHS Publication No. (SMA) 14-4884. Rockville, MD: Substance Abuse and Mental Health Services Administration. Supol, M., Satyen, L., Ghayour-Minaie, M., & Toumbourou, J. W. (2021). Effects of family violence exposure on adolescent academic achievement: A systematic review. Trauma, Violence, & Abuse, 22(5), 1042-1056. https://doi.org/10.1177/1524838019899486 Tabachnick, B. G., & Fidell, L. S. (2013). Using multivariate statistics (6th ed.). Pearson. Tabibi, J., Baker, L., Nonomura, R. (2020). Understanding resilience in children exposed to intimate partner violence – What we know from current literature Learning Network Brief 38. London, Ontario: Learning Network, Centre for Research & Education on Violence Against Women & Children. ISBN: 978-1-988412-43-6 92 Tamarit A., De la Barrera U., Schoeps K., Castro‐Calvo J., & Montoya‐Castilla I. (2023). Analyzing the role of resilience and life satisfaction as mediators of the impact of COVID-19 worries on mental health. Journal of Community Psychology, 51(1), 234–250. https://doi.org/10.1002/jcop.22900 Tedeschi, R. G., & Calhoun, L. G. (2004). Posttraumatic growth: Conceptual foundations and empirical evidence. Psychological Inquiry, 15(1), 1-18. https://doi.org/10.1207/s15327965pli1501_01 Trevillion K., Oram S., Feder G., & Howard L. M. (2012). Experiences of domestic violence and mental disorders: A systematic review and meta-analysis. PLOS ONE, 7(12), Article e51740. https://doi.org/10.1371/journal.pone.0051740 Ungar, M. (2013). Resilience, trauma, context, and culture. Trauma, Violence & Abuse, 14(3), 255–266. https://doi.org/10.1177/1524838013487805 Vezina J., & Hébert, M. (2007). Risk factors for victimization in romantic relationships of young women: A review of empirical studies and implications for prevention. Trauma, Violence, & Abuse, 8(1), 33–66. https://doi.org/10.1177/1524838006297029 Vu, N. L., Jouriles, E. N., McDonald, R., & Rosenfield, D. (2016). Children’s exposure to intimate partner violence: A meta-analysis of longitudinal associations with child adjustment problems. Clinical Psychology Review, 46, 25–33. https://doi.org/10.1016/j.cpr.2016.04.003 Wathen, C. N., & Macmillan, H. L. (2013). Children's exposure to intimate partner violence: Impacts and interventions. Paediatrics & Child Health, 18(8), 419–422. https://doi.org/10.1093/pch/18.8.419 93 Whittaker, T., & Stapleton, L.M. (2006). The performance of cross-validation indices used to select among competing covariance structure models under multivariate nonnormality conditions. Multivariate Behavioral Research, 41(3), 295–335. Wood, S. L., & Sommers, M. S. (2011). Consequences of intimate partner violence on child witnesses: A systematic review of the literature. Journal of Child and Adolescent Psychiatric Nursing: Official Publication of the Association of Child and Adolescent Psychiatric Nurses, Inc, 24(4), 223–236. https://doi.org/10.1111/j.1744-6171.2011.00302.x World Health Organization (WHO). (2013). Global and regional estimates of violence against women. World Health Organization. https://www.who.int/publications/i/item/9789241564625 World Health Organization. (2021). Violence against women prevalence estimates, 2018: Global, regional and national prevalence estimates for intimate partner violence against women and global and regional prevalence estimates for non-partner sexual violence against women. World Health Organization. Yick, A. G. (2008). A metasynthesis of qualitative findings on the role of spirituality and religiosity among culturally diverse domestic violence survivors. Qualitative Health Research, 18(9), 1289–1306. https://doi.org/10.1177/1049732308321772 Yule, K., Houston, J., & Grych, J. (2019). Resilience in children exposed to violence: A meta-analysis of protective factors across ecological contexts. Clinical Child and Family Psychology Review, 22(3), 406–431. https://doi.org/10.1007/s10567-019-00293-1 94 Zheng W., Huang Y., Fu Y. (2020). Mediating effects of psychological resilience on life satisfaction among older adults: A cross-sectional study in China. Health and Social Care in the Community, 28(4), 1323–1332. https://doi.org/10.1111/hsc.12965 95 Appendices Appendix A Initial Model 1: Structural Regression Model of Social Support and Resilience Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols. 1D21Resilience1 1Social SupportD1 11 1 1E21E3 E41 1Mental Health Physical HealthLife SatisfactionE51E6 E71 1Family IPV DisclosurePolice IPV DisclosureFriend IPV DisclosureCAS IPV DisclosureE111 96 Using the formula p* = p (p+1) /2, there are 7(8)/ 2= 28 non-redundant observations. There are 9 variances of exogenous variables and 1 direct relationship for a total of 10 parameter estimates (q). As the factor loadings are set to 1, they do not need to be estimated so they are not counted as parameter estimates to determine identification. Given that 28 pieces of information – 10 parameter estimates = 18; dfm ≥ 0. The additional requirements for identification of SR models are also met (i.e., there are at least two indicators per latent variable and the model is recursive). As such, Model 1 is overidentified. 97 Appendix B Initial Model 2: A Structural Regression Model of Spirituality and Resilience Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols. Mental Health Physical HealthLife Satisfaction1D21Resilience1 1E51E6 E71 1Religious IdentitySpiritualityD1 11Frequency of Formal PracticeFrequency of Informal PracticeImportance of Religion/Spirituality1 1E21E3 E41 11E11 98 Using the formula p* = p (p+1) /2, there are 7(8)/ 2= 28 non-redundant observations. There are 9 variances of exogenous variables, and 1 covariance for a total of 10 parameter estimates (q). As the factor loadings are set to 1, they do not need to be estimated so they are not counted as parameter estimates to determine identification. Given that 28 pieces of information – 10 parameter estimates = 18; dfm ≥ 0. The additional requirements for identification of SR models are also met (i.e., there are at least two indicators per latent variable and the model is recursive). As such, Model 2 is overidentified. 99 Appendix C Initial Model 3: A Structural Regression Model of Community Cohesion and Resilience Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols. Mental Health Physical HealthLife Satisfaction1D21Resilience1 1Community CohesionD1 11NeighbourTrustNeighbourHelpfulnessNeighbourConnection1 1E11E2 E31 1E41E5 E61 1 100 Using the formula p* = p (p+1) /2, there are 6(7)/ 2= 21 non-redundant observations. There are 8 variances of exogenous variables, and 1 covariance for a total of 9 parameter estimates (q). As the factor loadings are set to 1, they do not need to be estimated so they are not counted as parameter estimates to determine identification. Given that 21 pieces of information – 9 parameter estimates = 12; dfm ≥ 0. The additional requirements for identification of SR models are also met (i.e., there are at least two indicators per latent variable and the model is recursive). As such, Model 3 is overidentified. 101 Appendix D Initial Model 4: A Structural Regression Model of Social Support, Spirituality, Community Cohesion, and Resilience Note. The above model follows the McArdle-McDonald Reticular Action Model (RAM) model diagram symbols. Mental HealthPhysical HealthLife Satisfaction1D41Resilience1 1E121E13 E141 1Frequency of Formal PracticeSpirituality1Frequency of Informal PracticeImportance of ReligionD21Religious Identity1 11E71E61E51E81Community CohesionD3 11NeighbourTrustNeighbourHelpfulnessNeighbourConnection1 1E91E10 E111 1Social SupportD1111E31E2E111Family IPV DisclosurePolice IPV DisclosureFriend IPV DisclosureCAS IPV DisclosureE4111 102 Using the formula p* = p (p+1) /2, there are 14(15)/ 2= 105 non-redundant observations. There are 18 variances of exogenous variables, and 6 covariances for a total of 23 parameter estimates (q). As the factor loadings are set to 1, they do not need to be estimated so they are not counted as parameter estimates to determine identification. Given that 105 pieces of information – 23 parameter estimates = 82; dfm ≥ 0. The additional requirements for identification of SR models are also met (i.e., there are at least two indicators per latent variable and the model is recursive). As such, Model 4 is overidentified. 103 Vita auctoris NAME: Jenna Rose Emma Parsons PLACE OF BIRTH: Carbonear, NL YEAR OF BIRTH: 1998 EDUCATION: Carbonear Collegiate, Carbonear, NL, 2016 Dalhousie University, B.Sc., Halifax, NS, 2020 University of Toronto, M.Ed., Toronto, ON, 2021 |
|---|