Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Postestimation after asclogit

From   Simon <>
Subject   st: Postestimation after asclogit
Date   Wed, 05 Jun 2013 13:43:04 +0200

Dear Statalist,

I have a problem related to postestimation after -asclogit-

Using Stata 11, I am estimating an conditional logit model using -asclogit- with 11 alternative specific variables. The regression runs well and the output is expected. So far so good.

Following the example of Cameron and Trivedi (see ref below), I compute predicted probabilities with -predict pasclogit, pr- to compare the sample averaged predicted probabilities to the sample probabilities (observed proportions of each alternative), which gives an indicating of the precision of the estimated model. However, it seems that I am unable to get standard deviations regarding the predicted probabilities of the alternatives. Can anyone think of what may be wrong?

I also have some trouble in computing marginal effects for my variable of interest (g), using -estat mfx, varlist(g)-

asclogit y g $x, case(id) casevars() alt(alternatives) basealt() vce(robust) nocons nolog or
predict pasclogit, pr
table alternatives, contents(mean y mean pasclogit sd pasclogit)
estat mfx, varlist(g)

Table of probabilities,
Alternative    mean(Y)    mean(pasclogit)  sd(pasclogit)
1                  .35           .3597599                    0
2                 .0037234  .0108259                  0
3                 .0141844  .0126832                  0
4                 .0512411  .0398555                  0
5                 .0074468  .0042902                  0
6                 .0210993  .0220582                  0

Reference: Cameron & Trivedi (2009) Microeconometrics using Stata, chapter 15.5.7

Thanks in advance,

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index