I am looking to investigate the lagged effects of variables X and Z at t-1 in predicting a dichotomous categorical variable Y at t using the DSEM framework. I am wondering whether I should (or should not) include the autoregressive effects of X and Z (input lines 2-3 below) in the model? In what way does this change the interpretation of the spillover effects? I have run both models and get quite different results for some of the spillover effects.
%WITHIN% Y ON Y&1;! Carryover effect Y X ON X&1;! Carryover effect X Z ON Z&1;! Carryover effect Z Y ON X&1 Z&1;! Spillover effects of interest.