Re: [R] compare odds ratios

From: Smith, Daniel (DHS-DEODC-EHIB) <>
Date: Thu 29 Jun 2006 - 02:44:39 EST

You could try a test for heterogeneity of the odds ratios, usually part of a Mantel-Haenszel analysis, with each test as a stratum. Also, the field of meta-analysis has tests for heterogeneity of estimates. Both are covered in Rothman and Greenland's Modern Epidemiology (1998) text, Chapters 15 and 32 respectively.

Daniel Smith
Environmental Health Investigations Branch California Department of Health Services

> -----Original Message-----
> Message: 43
> Date: Tue, 27 Jun 2006 11:22:14 -0700 (PDT)
> From: array chip <>
> Subject: Re: [R] compare odds ratios
> To:
> Message-ID: <>
> Content-Type: text/plain; charset=iso-8859-1
> Hi dear all, I haven't heard any suggestions on how to
> tackle the problem in my previous email yet. I
> searched on google and was not getting any useful
> information yet. I did get someone from google groups
> suggesting Cockran Mantel Haenszel test with each
> subject as the stratum. But as far as I understand,
> CMH test is to test whether the common odds ratio
> (assuming odds ratios across stratum are equal) is
> equal to 1, not really the question I was asking which
> was whether the 2 odds ratios were equal (doesn't
> matter if they are equal to 1). Also, someone
> suggested loglinear regression, but I am not sure
> either how to set things up for my problem.
> One clarification to my original email: the 2
> diagnostic tests were performed on the set set of
> patients, the issue here how to test whether the 2
> odds ratios for the 2 diagnostic tests are equal.
> Here is a hypothetical dataset, for example:
> dat<-
cbind(disease=sample(c(rep(1,15),rep(0,20))),test1=sample(c(rep(1,11),re p(
> 0,24))),test2=sample(c(rep(1,14),rep(0,21))))
> Hope some statistical experts would guide me some
> directions. Many thanks
> ************************************** mailing list PLEASE do read the posting guide! Received on Thu Jun 29 03:03:08 2006

Archive maintained by Robert King, hosted by the discipline of statistics at the University of Newcastle, Australia.
Archive generated by hypermail 2.1.8, at Thu 29 Jun 2006 - 04:13:09 EST.

Mailing list information is available at Please read the posting guide before posting to the list.