Re: st: Predicted Probablities using Predict versus Prvalue for an MLOGIT

[Richard Williams <Richard.A.Williams.5@ND.edu>]

At 10:39 AM 12/10/2006, afremeth@csom.umn.edu wrote:
> I'm trying to calculate and manipulate the predicted probabilities for a
> multinomial logit with 3 outcomes for the dependent variable. When using
> predict p1 p2 p3 if e(sample), pr the summary stats (ie. the mean) on p1,
> p2, and p3 provide what would seem to be a reasonable result for predicted
> probabilities, however when I use prvalue to calculate the same thing it
> provides a result which would claim that the probability for one outcome is
> extremely and unrealistically low. What would explain this difference?
It would help to see your commands and output.  By default, prvalue 
plugs in the means for other variables.  This can be problematic if, 
say, you have interaction terms or dummy variables, where the values 
can't be mutually exclusive of each other, e.g. if you've got 
something like male*income, females will always be 0 on this 
term.  The 2nd edition of Long & Freese's book discusses these sorts 
of issues on pp. 425-427.


> Related to this is there way to examine the changes to predicted
> probabilities (ie. increasing by a standard deviation) by using the
> predict, pr command rather than a program like prchange.
> Thanks
prchange makes it so handy I am not sure why you would not want to 
use it.  But, other ways of plugging in hypothetical values are 
illustrated on pps. 7-10 of

http://www.nd.edu/~rwilliam/stats2/Logistic-Stata9.pdf

I think what Long and Freese do is create temporary cases with the 
desired values, and then use the predict command.  prchange automates 
that process but you can do it yourself, albeit somewhat more 
tediously.  There are also the prvalue and adjust commands.  Also, 
Long and Freese often show how you can use regular Stata commands to 
do the same thing as their commands.  I highly recommend the book if 
you don't have it already:

http://www.stata.com/bookstore/regmodcdvs.html



-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
FAX:    (574)288-4373
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW (personal):    http://www.nd.edu/~rwilliam
WWW (department):    http://www.nd.edu/~soc 
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/