st: Problem with sem and factor predictor variables

["Batterham, Alan" <A.Batterham@tees.ac.uk>]

I am using Stata 12 to analyse data from a parallel group randomised
controlled trial in a simple regression/ ANCOVA model in stata sem, to
take advantage of the mlmv method for missing data (full information
maximum likelihood). My problem relates to it not being possible in sem to
use the stata factor notation, as I am then unable to run the contrasts I
wish to run in postestimation. I have created dummy variables for factors
and my required interactions, but then cannot obtain the contrasts. I
thought that the new gsem in Stata 13 was my salvation, as gsem allows the
factor notation, but unfortunately it does not allow the mlmv option, so I
am between a rock and a hard place! On page 7 of intro 7 of the stata sem
documentation, it states

"constrast, pwcompare, and margins are postestimation commands of Stata.
All three can be used after gsem, and margins can be used after sem as
well. Even so, these
commands work best when you specify your model with Stata¹s
factor-variable notation; see
[SEM] intro 3. sem does not allow that notation, and so obtaining desired
results is a little more
work."

Can someone please help me with the "little more work" part, above? That
is, how do you get around the above issue to derive contrasts
postestimation? I can do everything I need to do in gsem using maximum
likelihood, or in standard regression, but then cannot use the mlmv
method, which is critical in this case. Anyone know why the mlmv option
was omitted from gsem?

Many thanks,

Alan Batterham. 


 


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/