I am running two kinds of multilevel models, 'type=twolevel' and 'type=twolevel random'. There are missing values on variables on both levels. When I run a 'type=twolevel' analysis and include variances in the model, only cases with missing values on all variables are excluded. So I assume that Mplus automatically uses the FIML estimator here. But when I run a 'type=twolevel random' analysis (again with variances in the model), cases with missings on x-variables are excluded. So I assume that Mplus does not use the FIML estimator in this case. This issue raises some questions: - Is it possible to use the FIML estimator with 'type=twolevel random'? - If yes: How can I use the FIML estimator with 'type=twolevel random'? - If no: Is there any other likelihood-based approach to handle missing values with 'type=twolevel random'?
Missing data theory is for observed dependent endogenous variables not for observed exogenous covariates. The model is estimated conditioned on the observed exogenous covariates. You can bring them into the model by mentioning their variances in the MODEL command. Then they will not be excluded but normality assumptions will be made about them.