Hi, I'm trying to run a MGA with categorical indicators for a latent dependent variable. there are 7 categories for each indicator. for one of the groups, there is apparently no case with a value of "6" on one of the indicators, and MPLUS gives me an error message because of this. Why is this a problem?
Also, I've gotten an error message telling me that I can't use delta parameterization with my model specification, and I should use theta parameterization. My hunch is that this is because I've tried to correlate categorical observed variables using the with statement. Is this the case?
The program is designed such that each group needs to have same values on categorical depdendent variables. You will need to collapse categories.
The Theta parameterization is not required when residual covariances are included in the model. It is likely that you have a categorical mediating variable. The conditions that make the Theta parameterization necessary are described in the Mplus User's Guide.
Hi, I'm running an MGA with 3 groups. when i specify my indicators as categorical, mplus kicks out 40 cases that are missing "on x-variables." i thought the missing data routine covered cases with missing data on independent variables...why would this be happening? Thanks!
Following is a quote from the Mplus User's Guide that addresses your concern:
"In all models, missingness is not allowed for the observed covariates because they are not part of the model. The outcomes are modeled conditional on the covariates and the covariates have no distributional assumption. Covariate missingness can be modeled if the covariates are explicitly brought into the model and given a distributional assumption."
In some models when all outcomes are continuous, there is no difference between x and y variables so that missingness is allowed on the x variables.
I have been estimating several path models with categorical outcome variables and have been using "Type is missing". I have noticed that in these models Mplus listwise deletes data missing on my predictor variables. This is consistent with what you have written on March 03, 2005 in quotes.
However, in path models with continuous outcome variables (using Type is missing), Mplus has not been listwise deleting data missing on my predictor variables (which are the same predictor variables as stated above). My predictor variables are binary and ordinal.
Does your last statement on March 2005 ("In some models...") explain why missing data on my predictor variables is not being deleted when I have continuous outcomes, and if so, could you explain the reasoning for this in more detail?
With TYPE=GENERAL MISSING and continuous outcomes, we use an all y model. This means that we make distributional assumptions about the exogenous variables. If you say TYPE=RANDOM MISSING, you will go down a path that estimates the model conditioned on the x variables.