Clint,
I think your confusion about the tests of homogeneity stems from what is
being tested. A test of homogeneity can be use to test whether the
exp-dis ORs differ over a third variable (effect modification) or
whether the ORs differ over several levels of an exposure
variable--regardless of whether a third variable is present.
-tabodds- allows you to examine the effect of an exposure that is
measured at more than 2 levels, e.g., 4x2; a 4-level exposure crossed
with a 2-level disease state. The real advantage of the -tabodds-
command is that the score test helps us to determine whether there is a
"dose-response" relationship. The test of homogeneity can be used to
detect a "threshold" of exposure. This test, though, still is based on
two variables. (-tabodds- allows adjustment for a third variable, but
not a test of interaction). The main thing to remember is to use
-tabodds- to examine a multi-level exposure.
The test of homogeneity in the -cc, by()- command is testing effect
modification. In this case, the test overlaps the -mhodds- command,
which does the same thing. We use this to address the question of
whether the exp-dis ORs are the same over strata of a third variable.
Obviously, if the test of homogeneity indicates that the ORs do differ
over strata of the third variable (i.e., the test is significant), using
a single statistic like the MH-odds to summarize the data may be quite
misleading.
pfv
______________________________
Paul F. Visintainer, PhD
Professor and Program Director
Health Quantitative Sciences
School of Public Health
New York Medical College
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of
[email protected]
Sent: Wednesday, July 07, 2004 2:31 PM
To: [email protected]
Subject: st: -tabodds- versus -cc-
Hello Statalisters --
I am using Intercooled Stata, version 8.2. I hope someone can
clarify the use of -tabodds- versus -cc- when determing whether
a Mantel-Haenszel (MH) odds ratio is appropriate. As I
understand it (and please feel free to correct me if my
understanding is misguided), the first step in determining
whether a MH odds ratio is appropriate is to generate odds
ratios for each group via -cc case_var exp_var, by(varname)-
then to follow this command with -tabodds case_var exp_var-
wherein I refer to the 'Test of Homogeneity (equal odds)' for
confirmation (or not) of whether a MH odds ratio is
appropriate. The Stata 7 reference manual (vol 1, pg 467),
however, indicates that I ought to refer only to the -cc-
output and its 'Test of Homogeneity (M-H)' in determining
whether a MH odds ratio is appropriate. Can anyone clarify?
As always, any suggestions are appreciated!
Many Thanks,
Clint Thompson
*
* 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/
*
* 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/
