topic Post hoc comparisons using Proc GLM in SAS Procedures

Post hoc comparisons using Proc GLM

deleted_user — Sun, 16 Aug 2009 08:17:34 GMT

Good Morning everyone,

I am new to SAS and have a question.

I have a dataset with one independent Group with 5 levels and one dependent variable Pol.

I am conducting a comparison of means using GLM with the following code:

proc glm data=mylib.'Sheet1$'n;
class Group;
model pol = Group;
lsmeans CDKstage / stderr;

The procedure works fine and shows a highly significant difference between the levels of Group p<.001. Now I would like to conduct a follow up analysis to determine which of the levels of group are different from the others. How do I do a post hoc comparison of means in Proc GLM.

Thank you!!

Re: Post hoc comparisons using Proc GLM

deleted_user — Sun, 16 Aug 2009 12:39:39 GMT

I believe I figure it out. I was able to conduct a Tukey pairwise comparison of means by simply adding a T to the lsmeans code as seen below. Because there have been a few views but no suggestion.maybe this will help others.

proc glm data=mylib.'Sheet1$'n;
class Group;
model pol = Group;
lsmeans CDKstage / T stderr;

Re: Post hoc comparisons using Proc GLM

Paige — Mon, 17 Aug 2009 12:35:37 GMT

The help for SAS 9.2 doesn't list T as an option in the LSMEANS statement.

It does list ADJUST=T as an option in the LSMEANS statement, which is NOT the TUKEY comparison of means. It is a simple t-test comparison of means.

Re: Post hoc comparisons using Proc GLM

statsplank — Mon, 17 Aug 2009 19:00:36 GMT

Hi Docnew,

Try the following:

proc glm data=mylib.'Sheet1$'n;
class Group;
model pol = Group;
lsmeans Group / pdiff adjust=tukey;
run;