Constraining the male and female parameter estimates to be equal resulted in a significant deterioration of fit for all models when compared to the models in which male and female parameters were estimated freely (Table 2). Consequently, male and female parameters were estimated separately in all further models. Consistent with the significant negative correlations between religiosity and problem alcohol use, reported above, it was not possible to drop the main effect of religiosity on the mean scores of the alcohol problem scores for males or females at either wave. Therefore, our results show that higher levels of religiosity are associated with lower mean levels of problem alcohol use.
