TTest by Class in Surveymeans?

[keherder]

Hello! I wish to conduct a ttest comparing means of a vairable between the two groups in the class variable, but it doesn't seem to be doing that. 

Here is my code:

proc surveymeans data=fus t;

cluster group;

var interest;

class period;

run;

Whether I take out or leave in the class statment, I get only a single t statistic for my interest variable. I would like a t test comparing the mean in terest between my two periods (in the class statement). Any suggestions how I could do this? I would like to account for the clusters, which is why I don't want to use a regular proc ttest. Thank you!

[PaigeMiller · Accepted solution]

I don't think SURVEYMEANS compares the means of the two groups, just the same way as PROC MEANS doesn't compare the means of two groups. It produces a t-test to see if the mean is equal to zero.

I think to get the t-test comparing differences between two means, you would have to use the linear model version of the t-test to compare two means

proc surveyreg data=fus;
cluster group;
class period;
model interest=period;
run;

Of course, all of this is untested; if you want tested code, provide (a portion of) the actual data as SAS data step code.

[keherder]

Thanks so much, this worked perfectly!