Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: predicted probabilities and confidence intervals for xtlogit


From   Thomas Gschwend <Thomas.Gschwend@mzes.uni-mannheim.de>
To   statalist@hsphsun2.harvard.edu
Subject   st: predicted probabilities and confidence intervals for xtlogit
Date   Thu, 01 Feb 2007 15:55:08 +0100

I am trying to make transparent the effect of a covariate by graphing predicted probabilities and their confidence intervals from a xtlogit model for several interesting scenarios. For people who run these models routinely this seems like a standard problem.
Across 8 years I am estimating the following random-effects model

xtlogit y A B C BC , re i(year)

with four covariates A, B, C, and BC. The forth covariate is an interaction term of B and C.
B is a dummy, A and C are continuous variables. I am interested in generating and subsequently graphing the effect of C on Y (for the entire range of C on the x-axis).
Searching the archives Joseph Coveney suggested once a potential way to approach this by generating extra observations that contain covariates with values for which I want the predictions (and setting the random effect to zero). Particularly I am interested in comparing the predictions in the cases where B=0 and B=1 for various values of C within its range given that A is set to its mean.

However, since my model also contains an interaction effect BC, the predicted standard errors or confidence intervals by -adjust- will be wrong. Is there another way to generate predicted values and correct standard errors that can be saved in a variable to be able to graph them later on?

Any help is greatly appreciated.
Thomas


*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/




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