topic Re: paired t-test for a group of variables in Statistical Procedures

paired t-test for a group of variables

fengyuwuzu — Wed, 25 May 2016 20:16:10 GMT

I have data in the following structure:

ID var1 var2 var3 when

1 10 15 20 After

1 9 14 18 Before

2 12 17 21 After

2 11 17 20 Before

3 10 16 23 After

3 11 15 22 Before

4 8 18 24 After

4 9 17 23 Before

...

If I want to do a paired t-test for each variable, it seems I need to create 3 datasets, one for each variable.

Is there simpler way to do it?

Thanks.

Re: paired t-test for a group of variables

ballardw — Wed, 25 May 2016 20:34:15 GMT

You would need to create one data set with the structure of:

Id BeforeVar1 AfterVar1 BeforeVar2 AfterVar2 BeforeVar3 AfterVar3.

Then the syntax would be:

Proc ttest data=have;

Paired BeforeVar1 * AfterVar1 BeforeVar2 * AfterVar2 BeforeVar3 * AfterVar3;

run;

Re: paired t-test for a group of variables

Rick_SAS — Wed, 25 May 2016 20:34:45 GMT

I think the simpler way to to combine each pair of rows. Basically moving from a "long form" of data that hass 3 measurements to a "wide form" that has 6 measurements.

If you strategically name the "before" and "after" variables, you can run all three analyses with a single PROC TTEST call:

data have2;
input ID 
      after_var1  after_var2 after_var3
      before_var1 before_var2 before_var3;
datalines;
1 10 15 20 9 14 18
2 12 17 21 11 17 20
3 10 16 23 11 15 22
4 8 18 24  9 17 23 
;

proc ttest data=have2;
paired (after:):(before:);
run;

Re: paired t-test for a group of variables

fengyuwuzu — Wed, 25 May 2016 21:00:06 GMT

Thank you, Rick!