I wonder if someone out there would be able to help me with the following problem.
We wish to look at the difference in proportions between the 'treat' and 'control' while adjusting for stratification variables(3 in this case). Randomised Controlled Trial. It is important that the result is given as a difference in proportions.
You can solve this by applying a multiple logistic regression model. This gives us odds ratios though.
You can use the Cochran Mantel-Haenszel test for example using PROC FREQ and this gives us odds ratios(OR) too. The riskdiff option seems to give us differences in proportion within each stratum. Have I missed something here?
I just wondered whether there was a non OR solution that would give us the difference in proportions including confidence intervals while adjusting for stratification variables.
Thanks for considering this problem.