Published Papers - Abstract 525

Agius P, Watson L, Taft A, Gunn J & Davey M. Women’s experience of violence and modelling data complexity: comparison of a typology approach with confirmatory factor analysis using structural equation modelling. BMC Public Health, ; :

Measuring complex phenomena with latent properties in quantitative research using secondary data can be a challenging process with issues such as the use of appropriate measurement methodology and construct validity often confronting the researcher. We use young cohort data from the Australian Longitudinal Study on Women’s Health (n=14,799) to provide examples of two analytical approaches to measuring women’s experience of violence. The first classified respondents into risk groups using a heuristic categorical composite measure and the second involved confirmatory factor analysis (CFA) to assess the fit of a congeneric latent factor represented by the items used to form the heuristic measure. Supporting the construct validity of both measures, ANOVA (heuristic measure) and structural modelling in AMOS analyses (CFA measure) showed negative associations between experience of violence and self-report physical and mental health. In concluding, we argue CFA provides a more rigorous method for testing the statistical fit of hypothesised measurement models, whereas classifying respondents into risk categories heuristically might support more clinically relevant analyses.