不好意思，应该怎么计算呢，我是菜鸟。治疗药物a有一个比率是80.4，药物b的比率是87.0。怎么计算这两个数值的估计差异呢，然后怎么计算这个差异的90%置信区间，用MH方法，多谢指教!!

你是用来做什么呢？如果是用来测试linear association的话，proc freq可以做这个检验

proc freq ...;
       table var1 * var2 / chisq riskdiff ;
       weight var3;
run;

不知道行不行的，试试吧

CMH was used to reweighting the proprotion in order to teesing out some confoundings, such as center effects. Sometimes we call that stratified analysis.
I was programming CMH analysis giving Wald conifidence interval. I also had code on CMH method in meta analysis, for examle, on Simpson's Paradox.
Please use and check the code in your own judgement. Jingju
http://blog.sina.com.cn/s/blog_a3a926360101d9k7.html