Re: st: Meta-analysis, Breslow-Day test for non-binary variable

[Steven Samuels <sjsamuels@gmail.com>]

The BD Test is, in your case,  a 2 d.f. of equal OR in the two cohorts; as such it is a test of the interaction between y and z in a logistic regression model.

***********************************
logistic  x  y##z,  coeflegend
test 2.y#1.z 3.y#1.z  
************************************

Steve

On Oct 19, 2011, at 8:09 AM, Berthold Hoppe wrote:

Dear all.

Using -mhodds x y, by(z)- I have analyzed the relation between -x- (0 or 1) and -y- (0 or 1 or 2) in two cohorts (-z-) to get pooled OR and M-H for homogeneity. 
In a review process I am now asked to provide a P for homogeneity by Breslow-Day instead of Mantel-Haenszel testing. Does anyone know if and how this could be calculated by STATA? 
As -y- is not binary -cc x y, by(z) bd- will not work.
I have tried to use -meta-, -metan- or -metacum- but a Breslow-Day test seem to be possible only if -y- is binary. 

Many thanks in advance,
Berthold
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/