I'm looking at indirect effects using WLS to get the indirect effects.
With y and inter are categorical I have
inter on age male; y on age male inter;
If I remove the "inter on age male" it makes a reasonable difference to the estimate for inter on y. Is this because inter is being corrected for age and male ? I don't seem to get the same problem with MLR.
When you remove inter ON age male, I suspect that you don't remove inter from the CATEGORICAL list. This means that the covariate inter is assumed to have an underlying normal latent variable whereas in the case where you have both equations in the model, only conditional normality given age and male for the latent variable behind inter is assumed. This is a less strong assumption.