I was having trouble getting a multiple indicators, multiple causes model to converge, so I simplified it to sem(X <- a b c) (X->d), which is the same as regr d a b c. The sem model still would not converge. It would produce the same parameters as the regr, after about 20 iterations, but would not converge. The log-likelihood from the regr and the sem were also much different. It seems like they should be the same. Any insights about what is going on and how to get the sem to converge (so I can then get back to the complete MIMC model) ? Thanks.