Can someone here suggest to me a suitable way to build a cross-lagged panel model with categorical variables on Stata MP 15.1? I built a cross-lagged panel model with categorical variables on Stata. Xs are binary variables, while Ys are continuous variables. They are repeated measurement across 5 waves. Equality constraints (i.e., a and b) were added as they have reciprocal associations. The causal directions of these variables are shown below. Array






In the first place, I used generalised response variables (i.e., family/link: Bernoulli / logit) to construct the variables of x, but I got errors when Stata displayed the GSEM results. Then, I changed the generalised response variables to observed variables (i.e., squares), and it ran smoothly. However, the effects of y on categorical x can’t be explained by the coefficients. That is, squares are for continuous variables, but Xs are binary variable. Alternatively, I should not use squares to construct Xs. What techniques should I apply to deal with this issue?