07-22-2014 01:09 PM
Wouldn't Fisher's Exact Test do what you need in PROC FREQ? At least as far as generating a probability statement about the "extremeness" of the observed proportions? Calculating the expected proportions would be more difficult, and if that is what you need, then you will probably need to explore PROC CATMOD.
07-23-2014 02:20 AM
My situation is that the value in each cell is known, for instance, at the very beginning I got the following table:
and suppose the control total of var1 and var2 are:
The target is to adjust the value of each cell of the var1 x var2 with the control total.