st: RE: RE: clogit or logistic for matched pairs

[VISINTAINER PAUL <PAUL_VISINTAINER@nymc.edu>]

Title: RE: clogit or logistic for matched pairs

My message should have said that -clogit- WOULDN'T work in the usual sense, because it requires the dependent variable to have at least one discordant pair in every cluster.

 

Paul

 

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of VISINTAINER PAUL
Sent: Thursday, November 06, 2003 1:58 PM
To: 'statalist@hsphsun2.harvard.edu'
Subject: st: RE: clogit or logistic for matched pairs

 

If I understand your description of the study, Michael, you don't have a case-control study, but rather a matched cohort study, (e.g., exposed twin vs. unexposed twin).  -clogit- would work for you in the usual sense, unless you regressed exposure status on disease.  You might want to try -xtlogit- which will take into account the clustering of your independent exposure variable. This will satisfy your need for a conditional approach.   Depending on your data and the covariates you include, you may actually get an outcome that is quite close to -logit- with a robust option.

Paul

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Michael Ingre
Sent: Thursday, November 06, 2003 11:01 AM
To: statalist@hsphsun2.harvard.edu
Subject: st: clogit or logistic for matched pairs

Dear statalist

I want to estimate the risk for an exposed twin to develop a symptom
relative an unexposed twin. I have 169 pairs of monozygotic twins that are
discordant on my exposure variable. The basic design is a case (exposed) -
control (unexposed) study. The outcome variable (symptom) is binary coded.

The most obvious method should be -clogit- however, I'm not satisfied with
the statistics.

Like McNemar, it excludes the ties between cases and controls and the odds
ratio returned describes the ratio of cases "winning" over controls.
However, I would like to estimate the ratio of a case to HAVE the symptom
(not "winning") compared to the control.

I have been criticized for using logistic regression with the cluster option
and robust variance estimator. My OR is said to be invalid.

Actually, the criticism is harder than that: My analyses is said to be
"completely meaningless".

. logistic dv iv , cluster(twinpairid)

Do you have any advice?

Thanks in advance

Michael Ingre

-----------------
PhD-student
Department of Psychology
Stockholm University &
National Institute for
Psychoscial Medicine
 

 

*
*   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/