Re: st: set reference dependent var category in longitudinal mlogit panel

[jpitblado@stata.com (Jeff Pitblado, StataCorp LP)]

Jason Bond <jbond@arg.org> is using -gsem- and wants to change the base
outcome for a multinomial logit model:

> I'm trying to estimate a longitudinal multinomial panel model using gsem. 
> The default dependent variable reference category is the first.is there an
> easy way to change this to a different category?  'baseoutcome(x)' is not
> allowed as an option and 'char depvar[omit] x' doesn't do anything.

Jason can use the 'b' operator to specify the base outcome for a multinomial
logit outcome variable in -gsem-.  The 'b' operator is part of factor
variables notation, see -help fvvarlist-.

In the following Stata log we use level 3 as the base for outcome 'insure'.
This example is based on '[SEM] example 37g'.

***** BEGIN:
. webuse gsem_sysdsn1
(Health insurance data)

. gsem (b3.insure <- i.nonwhite)

Iteration 0:   log likelihood = -556.59502  
Iteration 1:   log likelihood = -551.78935  
Iteration 2:   log likelihood = -551.78348  
Iteration 3:   log likelihood = -551.78348  

Generalized structural equation model             Number of obs   =        616
Log likelihood = -551.78348

------------------------------------------------------------------------------
             |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
1.insure <-  |
             |
  1.nonwhite |  -.3779586    .407589    -0.93   0.354    -1.176818    .4209011
       _cons |   1.941934   .1782185    10.90   0.000     1.592632    2.291236
-------------+----------------------------------------------------------------
2.insure <-  |
             |
  1.nonwhite |   .2828627   .3977302     0.71   0.477    -.4966742      1.0624
       _cons |   1.754019   .1805145     9.72   0.000     1.400217    2.107821
-------------+----------------------------------------------------------------
3.insure     |  (base outcome)
------------------------------------------------------------------------------
***** END:

--Jeff
jpitblado@stata.com
*
*   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/