Hi all. We are running a single-piece GCM on a longitudinal data with seven data-points and several time -variant and -invariant covariates. A WITH statement was used to get the correlation coefficients among the covariates. When we worked on the data without any missing values. By including or not including the WITH statement, there was no difference in both the model fit and point estimates (e.g., regression weights and standard errors). However, when we ran the models on the data imputed by Mplus, the point estimates were the same between two models (WITH vs. no WITH), but the model fit changed. With the WITH statement, the Chi-square increased (the # of free parameters increased but DF stayed the same). We checked the estimated covariances among covariates. They were exactly the same with the observed ones. Our question is, if the point estimates and estimated covariances are the same, where does the difference in chi-square come? Another observation is that Mplus automatically does multiple imputation when a "WITH" statement is included. Thank you in advance.