07-02-2015 08:56 PM
I need to run a pretty large multinomial comparison. I have a list of 2,000 different medicines, and I have computed their frequency "pre" and "post" (relative to a change in the experiment). I want to do a Pearson's test to see which of these medicines have significantly different proportional representations in the "post" sample.
So, my data look like:
Drug: Pre: Post:
A x1% y1%
B x2% y2%
If I had a smaller list -- say 3 medicines -- I would use:
PROC FREQ data=tmp
tables drug / chisq testp=( X Y Z )
But, I have around 2000 levels, and I don't want to type them out. I don't think proc freq can take the hypothesized values from inside the data, as I need. Is there another procedure that can handle such an input method? Or, is there a way I can batch/script this to produce the needed output?