|
|
Plots with multilevel complex sample ... |
|
Message/Author |
|
|
Mplus is crashing when it writes the .gh5 file, or when I try to access it after specifying any PLOT type with a multilevel complex sample analysis. Is this a bug, or doesn't it allow plots with that type of model? |
|
|
Please send the files and your license number to support@statmodel.com. I'm not sure what the reason is. |
|
|
Thanks, Linda, for letting me know about the problem with the graphics module. Until it's fixed, I'm hoping to obtain empirical Bayes estimates from the model I estimated, so that I can compute means for each of the subgroups. When I output FSCORES, I get an estimate for the outcome, but also a standard error. I assume that means these are not fixed predicted values. (When I create an SPSS file from them, for example, each of the subgroups have variance for the estimates.) However, in addition to the estimates for the outcome, I'm also getting an EBEst and standard error for the cross-level interaction between an L1 and an L2 variable that I computed with a DEFINE statement. The question is, then, how can I get fixed predicted values such as those I can compute from a model in SPSS or Stata, so I can obtain subgroup means of the predicted values? - bac |
|
|
The factor scores (FSCORES) are EB estimates. Because they are estimates they have SEs reflecting the estimation error/uncertainty. In other words, each person's factor score is an estimate and as such has a SE. You can still create subgroup means - with the usual caveat that estimated factor scores do not behave as true factor scores (this is why we have the SEM alernative). Some programs don't give the SEs, but that doesn't mean that the estimated scores are without estimation error. |
|
|
Sorry form my late response, Bengt - Somehow, I didn't get the email notice or it didn't get transfered from one computer the the other! Thanks for your explanation. This helps a lot. Best, Bruce |
|
Back to top |
|
|