topic Re: FIsher's exact pairwise comparison in SAS Procedures

FIsher's exact pairwise comparison

momi — Wed, 19 Apr 2017 15:27:36 GMT

Does anyone know how to do the pairwise comparison for Fisher's exact test?

Thanks,

Momi

Re: FIsher's exact pairwise comparison

ballardw — Wed, 19 Apr 2017 15:41:29 GMT

Provide some example data and the research question. Fisher has more than one test.

But I might guess that you want proc freq with an Exact statement and the Fisher option. You may need to make sure your data is in the correct structure.

Re: FIsher's exact pairwise comparison

momi — Wed, 19 Apr 2017 18:19:34 GMT

	group1	gorup2	goup3
A	7	6	11
B	3	3	6
C	15	1	7
D	0	0	4
E	13	8	14
F	10	2	3

Here is the data. All caterogires are nominal. According to the Fisher's exact test, there is a siginificant relashionship between the gorups and the categories A - F. Unfortunaly, I can not provide the actual measurement. (Does this matter?) I was wondering if it would be possible to the pari comparisons of these.

Thank you,

Momi

Re: FIsher's exact pairwise comparison

Ksharp — Thu, 20 Apr 2017 03:59:16 GMT

data have;
infile cards expandtabs truncover;
input name $ group1	group2	group3;
cards;
A	7	6	11
B	3	3	6
C	15	1	7
D	0	0	4
E	13	8	14
F	10	2	3
;
proc transpose data=have out=want ;
by name;
var group1-group3;
run;

proc freq data=want;
 weight col1;
 tables name*_name_/chisq;
 exact fisher /mc n=1000000;
run;

you can remove 'mc n=100000' to get the right fisher test,but that will cost you lots of time.

Monte Carlo Estimate for the Exact Test

Pr <= P 0.0375
99% Lower Conf Limit 0.0370
99% Upper Conf Limit 0.0380

therefore, category and group are not independent under alpha=0.05 .