Fig. 2

Predicted probabilities for being depressed by sex/gender and violence victimization, according to type of violence, primary aggressor, and frequency of victimization

Note: The figure shows the predicted probabilities for being depressed according to sex/gender, and: type of violence suffered (panel a), the primary aggressor (panel b), and frequency of aggressions (panel c). Victimization was defined as being the victim of at least one violent episode in the past 12 months. Depression was defined as having a PHQ-9 score > = 10. Predicted probabilities were estimated using the results of logit models that included as covariates: sex/gender, victimization (by type of violence, primary aggressor, and frequency, respectively), income quintile, race/ethnicity, age, region of residence, area of residence (urban/rural), marital status, cohabitation, employment, education, alcohol consumption, tobacco consumption, physical activity, health insurance and registered in primary healthcare services. In each case, predicted probabilities were estimated from a model that included interaction terms between the type of violence and sex/gender (panel a), between the aggressor and sex/gender (panel b), and between frequency and sex/gender (panel c). All other covariates were fixed at means. Results from analogous models, without the interaction terms, are depicted in columns 2 to 4 of Table A3 as odd ratios. All data were weighted according to the PNS survey design and survey weights