Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

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

Re: st: Binary Choice Model and fixed effects - interpreting the interaction effects?

From   Nick Cox <>
Subject   Re: st: Binary Choice Model and fixed effects - interpreting the interaction effects?
Date   Mon, 2 Apr 2012 11:23:50 +0100

As my own earlier post indicated, people interested in -inteff- should
download the later version from 4-3. This won't help your problem.

On Mon, Apr 2, 2012 at 11:19 AM, Benjamin Niug
<> wrote:
> Sorry for the shortcomings of the description of my problem:
> @Nick: It is  SJ4-2: st0063;
> @Maarten: Norton, Wang, Ai (2004), "Computing interaction effects and
> standard errors in logit and probit models", The Stata Journal 4,
> Number 2, pp. 154-167.
> Besides:
> @Maarten: Thanks for the odds ratio hint. Do you know of other
> solutions as well?
> Am 2. April 2012 12:10 schrieb Maarten Buis <>:
>> On Mon, Apr 2, 2012 at 11:57 AM, Benjamin Niug wrote:
>>> I want to estimate a binary choice model accunting for time-invariant
>>> fixed effects (I read I could use the -xtlogit- or -clogit- command).
>>> y_it = b_1*x_1_it*x_2_it+b_2*x_1_it + b_3*x_2_it
>>> However, I have included an interaction effect which I want to
>>> interpret correctly - as pointed out by Ai and Norton (2004) this is
>>> not trivial. They suggest to use a user written command called
>>> -inteff-. This command works well if -logit- is used, however, it does
>>> not work if -xtlogit- or -clogit- is used.
>> Please note that just author-year references are not appreciated on
>> this list. Please give the complete reference. This is discussed on
>> the Statalist FAQ. The logic is that this is a multi-disciplinary
>> list. Even if a citation is so famous within your
>> (sub-(sub-)discipline that author-year suffices, this is likely not to
>> be the case for the rest of the world. However, often many disciplines
>> will have independently faced (and solved) the same problem, and they
>> have something useful to say about the subject.
>> You have a double problem here: a) interpreting marginal effects of
>> interaction terms is hard, and b) interpreting marginal effects in
>> multi-level/panel/fixed effects models is hard. So the combination of
>> the two means that that is going to be very hard.
>> However, the solution is simple: don't do marginal effects but
>> interpret your coefficients in the natural metric of the model. In
>> this case the odds of success. Odds, odds ratios and ratios of odds
>> ratios have an undeserved reputation of being hard to interpret. You
>> can see an example of how easy that is here in:
>> M.L. Buis (2010) "Stata tip 87: Interpretation of interactions in
>> non-linear models", The Stata Journal, 10(2), pp. 305-308.

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index