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

st: Interaction terms in a logit model

From   "Daniel Schneider" <>
To   <>
Subject   st: Interaction terms in a logit model
Date   Fri, 18 Mar 2005 22:00:10 -0800

Dear List,

I have read the articles by Norton, Wang, Ai (2004) as well as their
more theoretical paper (Ai & Norton (2000)) and I am aware of other
literature describing the same problem. I think I understood the
theoretical problems and reasoning behind their approach, but
unfortunately I really have a hard time of really understanding what I
have to do when I use interaction terms in a logit regression. 

What I am trying to do is run a simple logit model with a simple
interaction term, something like this:

Y = b1X1 + b2X2 + b3X1X2 + controlling_variables

Where Y is a 0/1 variable.

I can run inteff, but I am just not able to interpret the results: What
does it mean that the effect is varying across the predicted
probability? My problem is that I would like to just do a -predict-
command, generate some simple graphs that show the different behavior of
the predicted probability depending on the manipulation of the two
variables (e.g., if X2 is a binary dummy, I would get two lines running
from left to right, when the yaxis is my predicted probability and the
xaxis is every value that X1 can take). To achieve this, I used
something like -prvalue- (or -prgen-) and just plotted the predicted
results later on. My understanding from the articles tells me that this
might be inaccurate, but I am now even confused about that.

So, if anyone can help me in interpreting the results of inteff beyond
saying something like "for some probabilities b3 is significant for
others it is not" - I can do that but I am not sure what that MEANS for
my model, because as far as I can see it I just want to show the readers
of my paper how a specific variable might have different influences
depending on a moderating variable. How can I show this in graphs or
data tables?
(I also read the DeLeire (2000) paper mentioned by Norton, Wang and Ai,
and I think this might be what I want to do, but I am not sure how to
achieve those results properly).

I really appreciate any help...

Daniel Schneider

*   For searches and help try:

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