Re: Fitting ordered probit models with endogenous covariates with Stata’s gsem command

Hi Dr. Canette
I am able to use a parallel method (introducing a latent variable with coefficient constrained to be 1) to get the covariance between the level one residuals in a multivariate multilevel model (with two ordinal dependent variables) using gsem i.e. the covariance was equal to the variance of the latent variable. However, I am unable to get the covariance between level one residuals using the same latent variable with same constraint for the empty model (with no covariates). Any advice would be helpful.