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]

Re: st: logit and mfx

From   Maarten buis <>
Subject   Re: st: logit and mfx
Date   Mon, 14 Feb 2011 08:08:40 +0000 (GMT)

--- On Mon, 14/2/11, Figueroa Armijos, Maria Augusta wrote:
> I have a question about -logit- and -mfx-. 
> I am running a logit model with three variants for the
> errors. First model is non-robust errors (-logit Y X1 X2-).
> Second model is with robust errors (-logit Y X1 X2,
> vce(robust)-). Third model is with clustered-robust errors
> (-logit Y X1 X2, vce(cluster fips)-).
> When I run -mfx- after each model, I get the same values of
> dydx (first differences) for non-robust, robust, and
> clustered-robust errors. What am I doing wrong?

You are doing nothing wrong. -robust- and -cluster- only 
influence the standard errors, it assumes that your model for
the effect of your explanatory variables on the probability is
correct. There is a nice discussion of what robust standard 
errors are (and what they are not) in the User's Guide. For a 
critical view on the consequences of this assumption see 
(Freedman 2006). My take on this is bit less gloomy: in my 
opinion the main problem is that name "robust" makes some users 
expect too much from it.

Hope this helps,

David A. Freedman (2006) `On The So-Called "Huber Sandwich 
Estimator" and "Robust Standard Errors"', The American 
Statistician, 60(4), pp. 299-302.

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen


*   For searches and help try:

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