Prep data for Fisher's exact test

[mariko5797]

I have data formatted as follows:

data have;
 input ID DOSENUM TRTN RESULT $ @@;
 cards;
 1	1	1	Y
 1	2	1	N
 1	3	1	N
 2	1	1	N
 2	2	1	N
 2	3	1	Y
 3	1	1	N
 3	2	1	Y
 4	1	2	N
 4	2	2	N
 4	3	2	N
 5	1	2	N
 5	2	2	Y
 6	1	2	N
 6	2	2	Y
 6	3	2	Y
 ;
run;

I want to find the p-value of Fisher's exact test comparing TRTN=1 to TRTN=2 for each dose. I am having a bit difficulty conceptualizing what needs to be tabled in PROC FREQ... I was thinking of doing the following, but am unsure if it gives me the p-value I want.

proc freq data= have; by DOSENUM; tables TRTN*RESULT/ fisher; run;

[svh]

This gives you p-values, but one assumption of the Fisher's exact test is that the subjects are independent. You appear to have repeated measurements and so a different test is likely necessary.

[Rick_SAS]

When you do a BY-group analysis of data, you need to make sure that the data are sorted by the BY-group variable. Before calling PROC FREQ, use the following call to PROC SORT:

proc sort data=have;
   by DOSENUM;
run;

[FreelanceReinh · Accepted solution]

Alternatively, add DOSENUM as a stratification variable to the TABLES statement:

tables DOSENUM*TRTN*RESULT/ fisher;

With that you can use the existing (unsorted) HAVE dataset, omit the BY statement and still obtain the same result: Fisher's exact test applied separately to the three dose groups.

However, from your description ("the p-value") I'm not sure if you want three p-values or rather a single p-value, maybe from an exact test for the common odds ratio, adjusting for DOSENUM, if appropriate for your data. This would be a kind of generalization of Fisher's exact test and could be requested by an EXACT statement (exact comor;).