|
|
Subpopulation Analysis with Complex S... |
|
Message/Author |
|
Emily Faust posted on Saturday, April 23, 2016 - 4:54 pm
|
|
|
I'd like to analyze national survey data (Add Health) using gender-stratified logistic regressions (i.e., each analysis run separately for males and females). One document that provided guidelines for analyzing this particular data set highlighted the importance of using special commands when conducting subpopulation level analyses; for example, the "subpop" command in Stata. Does Mplus have a similar command that is specific for subpopulation level analyses with complex survey data? |
|
|
See the SUBPOPULATION option in the user's guide. |
|
Emily Faust posted on Thursday, May 05, 2016 - 11:19 am
|
|
|
Thank you, Linda! Another question: Is is appropriate to use ML/MLR as a missing data estimator with categorical predictors and binary outcomes in logistic regression? I read that ML/MLR should not be used for data estimation with categorical variables in CFA, but I wasn't sure if the same thing applied to logistic regression. Furthermore, I've been trying (and struggling) to understand if there's a difference when using the Estimator command in Mplus for missing data estimation vs. specifying the model (i.e., using estimator=ml to request a logit model, rather than the default probit model). For example, could one use ml to request the logit model and combine that with WLSMV for missing data estimation? Any guidance would be very much appreciated! |
|
|
Maximum likelihood estimation can be used with categorical dependent variables for any model. No, you cannot combine estimators. With maximum likelihood, missing data estimation is FIML. With weighted least squares estimation, it is pairwise present. |
|
Back to top |
|
|