I wonder if it possible to include a mediation effect in multinomial logistic regression. I have a categorical (3 categories) outcome variable and four predictors (all continuous). I expect one of the predictors (X1) to mediate the relationship between the outcome variable and another predictor (X2). I also expect direct effects of X1 and X2 on the outcome.
Apparently, I could not see an option of running the regression in two steps (or blocks). Is there a way of including mediation in this analysis? I would be glad if you could inform me how it is possible.
PS: I am using PASW 18 for Mac.