Proc freq for 7x2 table, significant; how to pairwise

[fengyuwuzu]

I have age, gender (M,F) and sites (1,2,...7) columns, and I used proc freq to test (sites, gender ) and it showed p<0.001.

proc freq data= demo_sites ;
tables gender*sites_signup /chisq;
run;

But, how to do a pairwised chi-square test? I want to see which two groups are signigicant, and which two are not.

can I use proc multtest in this case?

A similar question for testing means of age in the 7 sites groups. I can use proc anova or proc glm to do a ANOVA test.

Is the solution to use proc multtest and contrast statement, like http://www.ats.ucla.edu/stat/sas/library/multtest.htm?

[Astounding]

This isn't an exact answer, but it may be helpful.  Similar to CHISQ, there is an option (CELLCHISQ?  may need to check my spelling) that adds to each cell its contribution to the overall chi-square.

[fengyuwuzu]

Thank you!! it is CELLCHI2.

With your hints I found one example:

http://support.sas.com/documentation/cdl/en/procstat/63104/HTML/default/viewer.htm#procstat_freq_sec...

Thank you again!

[SteveDenham · Accepted solution]

Regarding the testing of means, the LSMEANS statement in PROC GLM provides all of the options you need to compare all means in as many ways as you can think of, and adjust for the multiple comparisons using several methods.

proc glm data=yourdata;
class site;
model age=site;
lsmeans site/pdiff stderr cl adjust=sidak;
run;

This gives a Sidak adjustment for the 21 comparisons possible between the mean ages at the 7 sites.

Steve Denham

[fengyuwuzu]

@SteveDenham  Yes! This works excellently!