Multitests Post Significant Chi-Sq Omnibus Test in SAS

hlsmith

Less is more. Stay pure. Stay poor.
#1
I am hoping there is a basic and easy answer to my question.

I ran a 3x2 chi-square test, which was siginificant. I now would like to run the sub-2x2 tests, though I need to adjust alpha for the multiple tests. I could decrease my level of significance or apply a crude Bonferroni correction (e.g., 0.05/3), but many journals want the 0.05 the same and the p-values adjusted. I found a macro online that looked tedious and I saw the Westfall book had a chapter on this. Though, these pages were omitted in google books. Please let me know if you have any suggestions and if at all possible please supple sample code.

Thanks,


HS
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
I answered this by performing Proc Multtest. I included below some basic code that compares the three groups and also lists the different adjustment options. A source for examing these multiple test controlling for familywise error is: http://support.sas.com/kb/22/addl/fusion22950_1_multtest.pdf

proc multtest data=QI_SUP boot bon sid perm hoc stepbon ;
class group;
test fisher (Inapp);
contrast "Group1 V Group2" -1 1 1;
contrast "Group1 v Group3" -1 0 1;
contrast "Group2 v Group3" 0 -1 1;
run;