Re: st: predict after mlogit estimation

[Chiara Mussida <cmussida@gmail.com>]

Thanks Richard,
the 1st method seems more "precise". I will go for that one.
Nonetheless, the differnece between the two cases is not that much
relevant, on average. More precisely, some p's are almost equas, other
not with a discrepancies of (ar max) 0,018.



On 16/03/2012, Richard Williams <richardwilliams.ndu@gmail.com> wrote:
> At 04:09 AM 3/16/2012, Chiara Mussida wrote:
>>Thank you Richard,
>>this seems to be the solution. Anyway, is it possible to check whether
>>the results obtained for p1 p2 p3 are the ones I need (male
>>coefficients*female individual characteristics)?
>>Male coefs are in the stata output, whilst female characteristics in
>>my dataset. Should I compute the product between the coefs in the
>>output and the average individual characteristics of female?
>>
>>Thanks to you all
>>chiara
>
> If you want a 2nd opinion, try
>
> use "http://www.indiana.edu/~jslsoc/stata/spex_data/ordwarm2.dta";, clear
> mlogit warm yr89 white age ed prst if male == 1
> predict p1 p2 p3 p4 if male==0
> sum p1 p2 p3 p4
> margins if male == 0, noesample predict(outcome(1))
> margins if male == 0, noesample predict(outcome(2))
> margins if male == 0, noesample predict(outcome(3))
> margins if male == 0, noesample predict(outcome(4))
>
> Note that this is NOT the same as plugging in the average individual
> characteristics of females. If you wanted to do that, the commands would be
>
> margins if male == 0, noesample predict(outcome(1)) atmeans
> margins if male == 0, noesample predict(outcome(2)) atmeans
> margins if male == 0, noesample predict(outcome(3)) atmeans
> margins if male == 0, noesample predict(outcome(4)) atmeans
> * verify that female means were used by margins
> sum yr89 white age ed prst if male==0
>
> With the first approach, you are computing a prediction for each case
> and then averaging the predictions. With the 2nd approach, you are
> computing the predicted value for a person who had average values on
> all the independent variables. I generally prefer the first approach.
> It often doesn't make that much difference in practice, but sometimes it
> will.
>
> If you want yet a 3rd opinion, I suppose you could do the
> calculations for a few cases, or write your own gen command instead
> of using predict.
>
>
> -------------------------------------------
> Richard Williams, Notre Dame Dept of Sociology
> OFFICE: (574)631-6668, (574)631-6463
> HOME:   (574)289-5227
> EMAIL:  Richard.A.Williams.5@ND.Edu
> WWW:    http://www.nd.edu/~rwilliam
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>


-- 
Chiara Mussida
PhD candidate
Doctoral school of Economic Policy
Catholic University, Piacenza (Italy)
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/