Data question

[hein68]

I used code like this (see below) to pull the highest 5 scores out of a list of variables, and then I calculated the mean of the 5 highest scores.

array v{*} sd_1wk_1c sd_1wk_2c sd_1wk_3c sd_1wk_4c sd_1wk_5c sd_1wk_6c sd_1wk_7c sd_1wk_8c sd_1wk_9c sd_1wk_10c
 sd_1wk_11c sd_1wk_12c sd_1wk_13c sd_1wk_14c sd_1wk_15c sd_1wk_16c;
 
 array l{5};
    do i=1 to dim(l);
        l[i] = largest(i,of v(*));
    end;
 mdasisdhigh5_1wk=mean(of l1,l2,l3,l4,l5);

You can see the attachment to see what the resulting data look like.

Now I need to pull which 5 variables supplied the 5 highest values.  That will tell me which of the 16 symptoms were the highest for each subject.

How would I do that?

Thanks!

[PeterClemmensen]

How do you want the five contributing variables to be presented? As names in other variables or?

[hein68]

Names of the variables would be okay, but variable labels would be even better.

Thanks!

[Reeza]

Look at the VNAME and VINFO functions.

[Tom · Accepted solution]

Then you are going to have to get a different solution. The LARGEST() function just returns the values. Not where it came from.

Why not transpose the columns into rows to make it easier?

[hein68]

That helped.  I was able to figure it out.  Thanks!

[PeterClemmensen]

Do something like this and work from there. Makes the whole thing much easier 🙂

data have(drop=i);
   array v{*} sd_1wk_1c sd_1wk_2c sd_1wk_3c sd_1wk_4c sd_1wk_5c sd_1wk_6c sd_1wk_7c sd_1wk_8c sd_1wk_9c sd_1wk_10c
    sd_1wk_11c sd_1wk_12c sd_1wk_13c sd_1wk_14c sd_1wk_15c sd_1wk_16c;

   do _N_=1 to 10;
      do i=1 to dim(v);
         v[i]=rand('integer', 1, 10);
      end;
      output;
   end;
run;

data long(keep=n idx score var);
   set have;
   array v{*} sd_1wk:;
   n+1;
   do idx=1 to dim(v);
      score=v[idx];
      var=vname(v[idx]);
      output;
   end;
run;

proc sort data=long;
   by n descending score;
run;