Assessing threats to the valid measurement of violence against women constructs.

摘要

Establishing valid indicators of psychological constructs is fundamental to systematic and comparable research. In the emerging field of violence against women, the development and use of high quality instruments that measure constructs of psychological, physical, and sexual violence remains a crucial need. This study offers three steps toward meeting this need. First, it tests assumptions made about the structure of psychological, physical, and sexual aggression constructs. Second, it tests the invariance of these constructs in groups that differ by sample selection and measurement conditions. Third, it directly compares psychometric data from three measurement models to determine their relative efficacy for assessing the prevalence and frequency of violence against women. Data were obtained via individual interviews with 160 randomly selected men incarcerated in two state correctional facilities. Violence against women constructs were assessed with the Sexual Experience Survey (Koss & Oros, 1985), the Revised Conflict Tactics Scale (Straus, Hamby, Boney-McCoy & Sugarman, 1995), and the Severity of Violence Against Women Scales (Marshall, 1992).;Results indicate that reports of psychological and physical violence against women are prevalent in a random sample of criminal offenders, regardless of primary offense and criminal history. Sexual violence is reported less. Using items from the Conflict Tactics Scale Revised, confirmatory factor techniques analyzing prevalence data supported single-factor structures for psychological and physical aggression but no model adequately recovered an interpretable solution for sexual aggression items. More complicated measurement models for sexual aggression using items from the Sexual Experiences Survey were similarly non-conclusive. Results of metric and configural invariant models of the single-factor psychological and physical aggression suggest that these constructs are not invariant across sample selection defined by ethnicity. Moreover, measurement conditions that heighten the demand for social desirability render constructs of psychological and physical aggression unstable. Findings are discussed in the context of implications for existing literature and future psychometric needs in the field.