Item Response Theory Analysis of National Intimate Partner and Sexual Violence Survey Measures

• In Event: Results from the National Intimate Partner and Sexual Violence Survey (NISVS)

Abstract

The National Intimate Partner and Sexual Violence Survey (NISVS) includes lifetime and past-year measures of physical violence by intimate partners and sexual violence. The psychometric properties of these measures were examined using conventional methods (reliability analysis and factor analysis) and item response theory-based methods (two-parameter logistic item response models and analyses of differential item functioning). Results show that the NISVS questions are most informative for above average victimization levels. Results also show that items function differently for women and men. Differences are found at high to very high victimization levels. At those levels, some items have higher difficulty and discrimination for women, while others have higher difficulty and discrimination for men. Overall, these differences have little impact on summative victimization assessments. These assessments appear gender invariant, allowing us to make valid comparisons between women and men.

Author

• Andre B. Rosay, University of Alaska Anchorage