here is an answer but it is in R not stata, see the first link. There is an MPLUS package in R. The second link takes you to that documentation. best wishes.

http://stats.stackexchange.com/questions/51006/full-information-maximum-likelihood-for-missing-data-in-r

https://www.google.com/?gws_rd=ssl#q=Mplus+R

Thanks David! I know that its possible in R, but I think I can't make it with it, I never used R and it takes too much time to learn. :( 

Any Stata lovers here? 

As far as I know, you have two options in Stata:

1. Run mi (multiple imputation) commands and then run xtmixed. See http://www.stata.com/manuals14/mi.pdf

2. Run sem with an option method(mlmv); MLMV is an synonym for FIML in Stata. See http://www.stata.com/manuals14/semintro4.pdf#semintro4RemarksandexamplessemChoiceofestimationmethod However, sem does not allow factor variables. Generalized SEM (gsem) does allow factor variables but cannot use MLMV -- so if you have factor variables you can only work through using the mi command.

Thanks Alexander for this information, really helpful!! So sem could be an option ... unfortunately not for hierarchical data. Pity, but good to know that stata has limits here. Multiple imputation is not fiml, so its not working for me. 

**Regression using SEM - Full information maximuum likelihood**

sem (price

Franziska Cohen seems like you had the same challenge than me. Currently trying to overcome the missing variables in my dataset. I am wondering, if Multiple Imputation or Maximum Likelihood would be better? How have you solved it?

I am currently working with STATA and at the end I would like to perform a hierarchical regression ("hireg").

Many thanks in advance.

Elias Kurta

hi,

In the end, I calculated my models with MPLUS. SEM is supposed to work well (see above). In my field (education, psychology), ML would be the way; MI is not that known.

Hi Franziska,

many thanks for your answer.

Maximum Likelihood (FIML) sounds super interesting, nevertheless not really suitable to me, because I have to run a hierarchical regression afterwards (and to my understanding that is not possible with FIML In STATA).

Best,

Elias

Hi!

I am having similar hurdles, I use FIML to treat missing values, after using FIML in the regression model, interaction term and plot do not work anymore. Anyone has advice? Thank you!

Joy