experienced an accident but did not develop PTSD. The model specifies the probability of outcome Y = j (j = 1, 2, 3, 4), as a function of predisposition and potential confounders as follows: πj(xi)=Pr(Yi=j|xi)=exp(xi′βj)∑k=14exp(xi′βk), where xi is a vector of independent variables including predispositions and sex; and βj is the vector of corresponding coefficients for outcome j. For identification, the coefficients for the fourth outcome are set to zero, β4 = 0. The ratio of probabilities for outcome Y = 1 and Y = 3, i.e. PTSD after a sexual assault versus PTSD after an event of lower magnitude, for respondents with independent variable xi is therefore: π1(xi)π3(xi)=exp(xi′β1)exp(xi′β3). We test if this ratio is the same when a predisposition is present versus when it is absent. Using MDD as an example, when MDD = 1, controlling for sex, the above ratio is: π1(xi)π3(xi)|MDD=1=exp(β10+β11+β12SEX)exp(β30+β31+β32SEX), whereas when MDD = 0, the ratio is: π1(xi)π3(xi)|MDD=0=exp(β10+β12SEX)exp(β30+β32SEX). The ratio of the above two expressions is exp(β11)/exp(β31), which is conventionally called a relative risk ratio (RRR) or odds ratio (although neither term is strictly accurate). The assumption that predispositions have a stronger influence on the risk of PTSD when events are less severe can be
