Re: st: Interactions using xtlogit

[Maarten Buis <maartenlbuis@gmail.com>]

On Thu, May 10, 2012 at 12:33 AM, Sturrock, Hugh wrote:
> I am running a random effects model to model infection data (IsPos) using the xtlogit command and want to include an interaction between 2 categorical variables, one with 3 groups (Dist) and the other with 2 groups (Local). I'd ideally like a single p-value to tell me whether the association between IsPos and Dist is modified by Local.
>
> Assuming ID is the household level ID I'm clustering by, if I run the following
>
> xi:xtlogit IsPos i.Dist*i.Local, i(ID) or
>
> I get two interaction terms DistxLoc_2_2 and DistxLoc_3_2. To see if there is an overall interaction occurring I've used the command
>
> test DistxLoc_2_2 DistxLoc_3_2
>
> My question is, is this the correct approach?

Yes, as long as you interpret the interaction effects as ratios of
odds ratios. At first glance some people find that complicated, but if
you interpret the model in a step by step basis, starting with the
baseline odds, continuing to the odds ratio, and than the ratio of
odds ratios it does become as clear as is possible with a complicated
concept like an interaction effect. An example of such an analysis is
given here:

M.L. Buis (2010) "Stata tip 87: Interpretation of interactions in
non-linear models", The Stata Journal, 10(2), pp. 305-308.
<http://www.maartenbuis.nl/publications/interactions.html>

> Also, when do you use # instead of * to indicate an interaction?

Always when the program you use allows it (there are some older
programs around, including some of my own, that do not (yet) allow
factor variables.).

> Also, if I wanted to make some other pairwise comparisons between interaction groups (i.e. DistxLoc_1_1 with DistxLoc_1_2) how would I do it?

-help margins_pwcompare-

Hope this helps,
Maarten

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


http://www.maartenbuis.nl
--------------------------

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/