Statalist


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

Re: st: t: Test for Interaction-Meta analysis program


From   tiago.pereira@incor.usp.br
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: t: Test for Interaction-Meta analysis program
Date   Sat, 22 Mar 2008 23:45:12 -0300 (BRT)

To your question: if I understood it correctly, I would say YES under a
meta-analysis view. I believe this conjecture is correct as long as you
can express your statistic in the weighted form.

A test for interaction (heterogeneity or trend ) aims to determine whether
or not there is evidence of difference in effect sizes between subgroups
(under H0, it is expected that the effect sizes (ES) are the same across
strata: ES1=ES2=.....=ESn).

Tom Trikalinos, a frequent Statalist contributor, is an expert in
meta-analysis and will be able to clarify better this issue.

Hope this helps.
Tiago


-----------------


Thanks for the reply.  Is test for heterogeneity between subgroups the
same as test for interaction?








Sripal,

You have several strata, say, N 2x2 tables (or effect sizes and their
variances/standard errors), you can test for interaction (assuming that
interaction in this case means lack of homogeneity) using either the
Cochran Q-Statistic or the Breslow-Day test.

-metan- calculates automatically the Q-statistic, whereas the Breslow-Day
test requires the option breslow as well as raw data (complete 2x2
tables).  Assuming you have raw data (2x2 tables) as a b c d (four
collumns), type, for instance:

metan a b c d, or random breslow

or

metan a b c d, rr fixedi

see the several options you have on the help file.

Tiago






*
*   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