CE Carey posted on Thursday, August 13, 2015 - 1:23 pm
I am trying to test a mediational model with a continuous observed independent variable X, a continuous observed mediator M, and a binary observed outcome Y. I also have a few uncorrelated binary and continuous covariates of no interest (covs1-5) that I would like to regress on the mediator and outcome. I have specified the model as follows:
Y on M (b1); Y on X (c); M on X (a1); M Y on covs1-5;
However, when I try to run the model, it will not converge. Based on some troubleshooting, I believe that this is due to the regression of the binary outcome on the continuous covariates of no interest: when I no longer declare Y as categorical, the model runs fine. I assume that this is not a viable solution though, since Y is indeed categorical, and not declaring it as such would change the distributional assumptions the variable. Please let me know what, if anything, would allow for the proper specification of this model.