topic Re: paired t-test - differences by group in SAS Procedures

paired t-test - differences by group

LucyB — Thu, 09 Mar 2017 21:36:45 GMT

I have pre/post data for one group - outcome is a score. I am interested to see if the scores vary by a covariate like gender. If I use the by command in the ttest procedure, it does not give me a significant p value for differences by gender. how can i go about doing this?

Re: paired t-test - differences by group

Reeza — Thu, 09 Mar 2017 22:09:12 GMT

I think you need a PROC ANOVA not a T-Test.

Re: paired t-test - differences by group

Reeza — Thu, 09 Mar 2017 22:11:14 GMT

Also, BY statements splits an analysis for each group. It's the equivalent of doing some repeatedly for each group but they're considered individually. Analyses do not cross between BY groups.

Re: paired t-test - differences by group

LucyB — Thu, 09 Mar 2017 23:35:59 GMT

So would be outcome variable be the difference? And my independent would be gender, but do I need to adjust for baseline (pre) scores as well?

Re: paired t-test - differences by group

Reeza — Fri, 10 Mar 2017 02:33:33 GMT

Pre-calculate the differences and test if the difference is different between the groups and not equal to 0.

ie Mean from Group A = Mean from Group B = Mean from Group = 0

where mean = post-pre score