Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: marginal effects after nlogit and clogit


From   May Boggess <mboggess@stata.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: marginal effects after nlogit and clogit
Date   23 Aug 2004 10:21:40 -0500

On Monday, Sascha asked:
> 
> my understanding is that -mfx- cannot be used to obtain marginal effects
> after -clogit- or -nlogit-
> (see www.stata.co.uk/support/faqs/stat/mfx_ologit.html)
> 

This FAQ is for Stata version 7 only. At the top of the page there
is a message to that effect (we leave the version 7 FAQ's there for
version 7 users). 

> How does one go about calculating marginal effects then after estimating
> -nlogit-?
> 

Please look at this FAQ:

 http://www.stata.co.uk/support/faqs/stat/mfx_unsuit.html
 
This explains in detail why some prediction functions are suitable and
some are not. 

In short, -mfx- calculates the partial derivatives of the predict
function specified, and so it needs to be a function of only the
covariates and the coefficients of the model. If the prediction function
is a function of some other quantity, then -mfx- cannot possibly know
how to take the derivative. So -mfx- does its best to check the predict
function and it refuses to calcuate the derivatives if there is any risk
it can't do it correctly.

There is a detailed example in the FAQ of how these checks are done
using -clogit-.

The only prediction functions I found that are suitable after -nlogit-
are xbb, xb1 and xb2. But these, I don't think, are terribly informative
(if you are seeking marginal effects not elasticities), because these
are linear functions of the covariates and coefficients, and so the
derivatives will just be the coefficients of the model. 

--May
mmb@stata.com



*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index