Gsem with the method option 'mlmv' in Stata 15 ?

[Sule Yaylaci]

Dear Statalist users,
I am trying to do exploratory factor analysis(EFA) with 8 ordinal variables and each variable has about 10-15% missing values. I am using Stata 14.
MPlus has an option of wlsmv for EFA for such variables, but as far as I understand, 'factormat' command in Stata does not offer any full information maximum likelihood method.
I came across this link which describes doing efa within a cfa framework: https://stats.idre.ucla.edu/stata/fa...work-stata-12/
However, the link uses sem and I need to use gsem, and gsem does not have the option 'mlmv' unlike sem in Stata 14. I was wondering if Stata 15 added that option.
I don't want to use 'sem' because of the ordinal nature of my variables, and they violate the multivariate normality assumption.
Gsem seems to handle missing values much better than 'list-wise deletion' but mlmv could use more information, and I don't know if Stata 15 has that feature or a similar feature to make it worthwhile to upgrade.
Thanks for your help in advance.
Sule

[Clyde Schechter]

As far as I know, the -mlmv- method is not available for -gsem-.

[Richard Williams]

I'd like to see mlmv, or something like it, added to many commands, e.g. regress. But, as far as I know, mplus is the only program that supports it with gsem-type models.

[Sule Yaylaci]

Thanks much, Clyde and Richard.