I think there should be a correction. Chisq , Mantel-Haenszel tests mainly test the association between variables rather than test the trend in proportions over time. This was the same question I was looking for the answer these days. Then I found Cochran-Armitage Test for the trend in binomial proportion.
According to the natural of binomial distribution, the sampling scheme for Cocharan-Armitage only works on 2 x r tables. Therefore in SAS
proc freq data=ur_data;
table answer*year/TREND; /* provide Cocharan-Armitage test for trend*/
weight count_in_ur_data;
run;
if test statistic <= 0, it means decreasing trend, otherwise increasing.
Still for the trend in proportion, Is there any test that I can compare whether the trends in proportion over time between two groups are parallel? I know regression can work to compare the slopes, but what about statistical test ?
扫码加好友，拉您进群
全部版块
我的主页
收藏
