Rebecca Weston

A Brief Guide to Structural Equation Modeling

To complement recent articles in this journal on structural equation modeling (SEM) practice and principles by Martens and by Quintana and Maxwell, respectively, the authors offer a consumer’s guide to SEM. Using an example derived from theory and research on vocational psychology, the authors outline six steps in SEM: model specification, identification, data preparation and screening, estimation, evaluation of fit, and modification. In addition, the authors summarize the debates surrounding some aspects of SEM (e.g., acceptable sample size, fit indices), with recommendations for application. They also discuss the need for considering and testing alternative models and present an example, with details on determining whether alternative models result in a significant improvement in fit to the observed data.

PTSD symptoms among men and women survivors of intimate partner violence: the role of risk and protective factors.

The purpose of this cross-sectional analysis of the National Violence Against Women Survey was to characterize current symptoms of posttraumatic stress disorder (PTSD) among 185 men and 369 women survivors of intimate partner violence (IPV). In this subsample, 24% of women and 20% of men had current moderate-to-severe PTSD symptoms. PTSD scores were higher for women than men. Protective factors that appear to increase resiliency of survivors were higher education and income, being currently married, and reporting that IPV had stopped. Higher physical or psychological IPV scores, current depressive symptoms, and the survivor having left the relationship at least once were associated with risk of moderate-to-severe symptoms of PTSD. Protective factors may be used to boost resiliency of IPV survivors and reduce PTSD symptoms.

Differing Effects of Partner and Nonpartner Sexual Assault on Women's Mental Health:

This study contrasted the effects of intimate partner and nonpartner sexual assault on womens mental health among a sample (N = 835) of low-income, ethnically diverse community women. Compared to sexual assault by a previous partner or by a non-intimate partner, sexual assault by a current partner was the strongest predictor of PTSD, stress, and dissociation. Non–intimate partner sexual assault was only a significant predictor of PTSD and only for African American women. These findings suggest that the victim-offender relationship is important when considering the impact of sexual assault. Specifically, sexual assault perpetrated by an intimate partner may be especially traumatic.

Childhood and adolescent sexual abuse of community women: mediated effects on psychological distress and social relationships.

UNLABELLED Possible mediators of sexual abuse severity were tested on the basis of D. Finkelhor and A. Brownes (1985) traumagenic dynamics model with 178 low-income African American, European American, and Mexican American community women interviewed for Project HOW Health Outcomes of Women. This subsample reported contact sexual abuse before the age of 18 years. Severity was level of force, number of perpetrators, relationship to perpetrator, and age at first assault. As expected, structural equation modeling showed powerlessness, and stigmatization largely mediated the effects of sexual abuse severity on womens psychological distress in adulthood. Powerlessness also mediated the effects of severity on maladaptive social relationships. The expected path from betrayal to relationships was nonsignificant. Overall, the results support extension of D. Finkelhor and A. Brownes model. Possible interventions are addressed.

Toward Ethnically Specific Models of Employment, Public Assistance, and Victimization

Among 836 low-income women, those receiving Aid to Families With Dependent Children (AFDC) or food stamps had experienced more coercive sexual assault, abuse by past partners, psychological abuse by current partners, and types of victimization than women not receiving assistance. The two groups of women were equally likely to endure threats or violence from current partners. African Americans and European Americans were more likely to have been victimized than Mexican Americans. European Americans reported more severe victimization except current partner violence. Multiple regressions on employment and assistance showed victimization predictors that varied by ethnicity. The effects of abuse by current partners were limited and are likely to be indirect.

Physical and mental health outcomes of women in nonviolent, unilaterally violent, and mutually violent relationships.

Despite equivocal findings on whether men or women are more violent, the negative impact of violence is greatest for women. To determine how gender asymmetry in perpetration affects women’s health status, we conducted a study in two phases with 835 African American, Euro-American, and Mexican American low-income women in Project HOW: Health Outcomes of Women. In Phase 1, we used severity and frequency of women’s and male partners’ violence to create six groups: nonviolent (NV), uni-directional male (UM) perpetrator, uni-directional female (UF) perpetrator and, when both partners were violent, symmetrical (SYM), male primary perpetrator (MPP), and female primary perpetrator (FPP). The MPP group sustained the most threats, violence, sexual aggression, and psychological abuse. They also reported the most fear. Injury was highest in the MPP and FPP groups. In Phase 2, we examined group differences in women’s health status over time for 535 participants, who completed five annual interviews. Surprisingly, women’s health in the MPP and FPP violence groups was similar and generally worse than if violence was uni-directional.

Path analytic examination of a cognitive model of PTSD

Ehlers and Clark (2000) developed a cognitive model of posttraumatic stress disorder (PTSD) symptom maintenance which implicated the role of posttraumatic cognitions and aspects of the trauma memory in maintaining symptoms via an increased sense of current threat. The aim of the current study was to empirically test a variant of this model using path analysis. Participants in the current study were 514 undergraduates at a midwestern university who reported experiencing at least one traumatic event. Path analyses examined various models of the possible relationships between ones posttraumatic cognitions and the centrality of the traumatic event to the sense of self (considered an aspect of memory integration) in predicting current level of PTSD symptoms. Results indicate that both event centrality and posttraumatic cognitions are unique and independent predictors of current symptom level. Overall, the results of this study support aspects of Ehlers and Clarks cognitive model of PTSD; cognitive appraisals of the self and centrality of the event were highly related to levels of distress. However, the current study suggests that overly integrated trauma memories may lead to greater distress and not poorly integrated ones as suggested by Ehlers and Clark.

Is Plagiarism a Forerunner of Other Deviance? Imagined Futures of Academically Dishonest Students

This study explored the relationship of current incidences of academic dishonesty with future norm/rule-violating behavior. Data were collected from 154 college students enrolled in introductory and upper-level psychology students at a large Midwest public university who received credit for participating. The sample included students from many different majors and all years of study. Participants completed a self-report survey that included a measure of Academic Dishonesty (including three subscales: Self-Dishonest, Social Falsifying, and Plagiarism) and an Imagined Futures Scale (five subscales that included Norm/Rule Violating, Physically Threatening, Culturally Diverse, Emotionally Distressing, and Agentic Futures). Correlation analyses indicated a significant positive relationship between all three Academic Dishonesty subscales and an imagined norm/rule-violating future. Further, regression analyses revealed social falsifying as being significantly predictive of a norm/ rule-violating future. Suggestions are made alerting educators to the importance of monitoring and discouraging academic dishonesty as it may lead to rule-violating behavior in the future.

Insecure Attachment Mediates Effects of Partners’ Emotional Abuse and Violence on Women’s Relationship Quality

Men’s emotional abuse and violence have a broad and pervasive impact on women that may include long-term effects on women’s attachment and relationship quality. In this longitudinal study, women’s Wave 6 ratings of their insecure attachment were hypothesized to mediate the relationship between partners’ Wave 5 abuse (emotional and physical) and Wave 6 relationship quality, with differences in associations by women’s Wave 5 self-classification as secure or insecure. Mediation was tested with data from a sample of 574 African American, Euro-American, and Mexican American community women who had completed at least three waves of a six wave study. Differences occurred in the final structural equation models by women’s Wave 5 attachment style, with direct paths from emotional abuse to insecure attachment and from violence to relationship quality for both groups, but direct effects of violence on relationship quality only for insecurely attached women.

Does Men's Positivity Moderate or Mediate the Effects of their Abuse on Women's Relationship Quality?

Taylors (1991) mobilization-minimization theory of coping with stress provides one perspective to understand counterintuitive findings from research on abuse of women (e.g., many abused women love their partner and report relational satisfaction). Extension of her theory led to the hypothesis that mens affection and caring behavior mediates the effects of their abuse on womens evaluative relational judgments (relational well-being, satisfaction, happiness, and commitment). Using an abused subsample from a larger study of low-income, ethnically diverse women, we tested whether mens positivity (caring behavior, relational satisfaction, and happiness) moderates their abuse (threats and acts of violence and sexual aggression) as suggested by Lloyd (1996) or functions as a mediator. The results of structural equation modeling with 717 African-American, Euro-American, and Mexican-American women indicated that Mens Positivity did not moderate the effects of Mens Abuse on Relational Quality. Partial and complete mediation provided similar fits, with the full mediation model accepted as the most parsimonious. This held for women who had sustained severe, potentially life-threatening violence. Neither ethnicity nor relational type (dating, cohabiting, married) functioned as moderators. Thus, when women are abused they apparently give more weight to their perceptions of the positive aspects of their partner when evaluating the quality of their relationship. Limitations of the study as well as implications for therapy and future research are discussed.