I am currently performing some simulations of Thurstonian-IRT models and, for certain simulation conditions, have considerable problems getting Mplus to converge. My inituition tells me this could be because of (quasi) complete separation in parts of the model, but I am not really sure. The main problem I have is that for non-converging models, Mplus is not willing for give me any factor scores, which I need to evaluate my simulations. The errors and warnings I see are as follows:
WARNING: THE BIVARIATE TABLE OF I25I26 AND I13I14 HAS AN EMPTY CELL. (and a few others of that kind)
NO CONVERGENCE. NUMBER OF ITERATIONS EXCEEDED. (I currently use 3000 iterations)
FACTOR SCORES WILL NOT BE COMPUTED DUE TO NONCONVERGENCE OR NONIDENTIFIED MODEL.
I would appreciate any help to either get the model to indicate convergence or report the factor scores regardless of convergence.