paired ttests and means

[starz4ever2007]

is there any way to do proc ttest on multiple variables. for instance:

I have a dataset with variables :

subjid include _1_abc_q1   _1_abc_q2  _1_abc_q3_8_abc_q1 _8_abc_q2  _8_abc_q3

   1        1            6                    4                  3               5                4                  3

   2        1            6                    1                  3               6                4                 5

and I want to do paired ttest for _1_abc_q1*_8_abc_q1 

                                                   _1_abc_q2*_8_abc_q2

                                                   _1_abc_q3*_8_abc_q3 

also can I get the means separately for each of the variables in the paired ttest in addition to the difference of the means (e.g., mean of _1_abc_q1 and mean of *_8_abc_q1)...keeping in mind that just doing a proc means on each variable would include observations not present in the paired ttest (due to missing observations in one of variables).

I do have another dataset that sets up the same data in another way so that the 1 and 8s are variables and the q1-q3 are variables if doing a ttest/means would be easier/ possible this way:

subjid   include        wk     q1    q2    q3

    1            1             1       6       4     3

    2            1              1       6      1     3

    1            1              8       5      4     3

    2           1               8       6       4    5

[starz4ever2007]

I was able to get the t tests to run using:

proc ttest data=data1;

paired ( _1_abc: ) : ( _8_abc: );

run;

but still unable to figure out how to get the separate means....

[art297 · Accepted solution]

I don't know of an easy way. You could always use something like:

data data1;
  input subjid include _1_abc_q1   _1_abc_q2  _1_abc_q3 _8_abc_q1 _8_abc_q2  _8_abc_q3;
  cards;
1     1     6    4     3    5    4   3
2     1     6    1     3    6    4   5
;

proc ttest data=data1;
  paired _1_abc_q1*_8_abc_q1 _1_abc_q2*_8_abc_q2 _1_abc_q3*_8_abc_q3;
run;

proc means data=data1 (where=((not missing(_1_abc_q1)) and (not missing(_8_abc_q1))));
  var _1_abc_q1 _8_abc_q1;
run;

proc means data=data1 (where=((not missing(_1_abc_q2)) and (not missing(_8_abc_q2))));
  var _1_abc_q2 _8_abc_q2;
run;

proc means data=data1 (where=((not missing(_1_abc_q3)) and (not missing(_8_abc_q3))));
  var _1_abc_q3 _8_abc_q3;
run;

Art, CEO, AnalystFinder.com

[starz4ever2007]

thanks 🙂

yea I can't seem to find an easier way than that either.