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: marginal effect after xtlogit model

From   Richard Williams <>
Subject   Re: st: marginal effect after xtlogit model
Date   Tue, 25 Oct 2011 22:49:40 -0500

At 09:15 PM 10/25/2011, ramesh wrote:
I am using xtlogit model, which is in the form of: y=
b0+b1.x1+b2.x1*x1+b3.x3 +b4.x4). I have one squared term (x1*x1) and two of
my independent variables (x3 and x4) are binary and after using the 'xtlogit
y x1 c.x1#c.x1 x3 x3' followed by 'margins, dydx(*) atmeans' I got the same
beta coefficient as marginal effect, which is not correct. Is there any idea
to correctly specify the marginal effect?

For binary and other categorical variables, you should explicitly indicate that the variable is categorical by using the i. notation. Otherwise margins will assume the variable is continuous. In this case,

xtlogit y x1 c.x1#c.x1 i.x2 i.x3

I also prefer to not use the atmeans option, but some people disagree about this and it may not make much difference in practice.

I think your main problem, though, is that the default predict option for xtlogit is (I think) xb. See help xtlogit postestimation##predict and figure out what option you want to use instead, and specify it as part of the -predict- option on margins.

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

*   For searches and help try:

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