What seems simple is complicated.

[tmccombs27]

I have an order transactions data set with one million records.  Among the list of order records and variables, there is a GROUP variable which is a classifier and an ITEMS variable which is a numerical integer.  My goal is to make a table that displays the top 10 GROUPs with the highest mean ITEMS.  I will try to give you an example data set:

order_id, group_id, items;

1, A, 6;

2, D, 12;

3, B, 15;

4, A, 10;

5, C, 2;

6, B, 14;

7, F, 4;

8, B, 9;

9, A, 24;

10, D, 12;

I imagine I need to use a combination of PROC MEANS, PROC SORT, and PROC PRINT, but I have tried these in combinations with no success. 

[PaigeMiller · Accepted solution]

proc summary data=have nway;
    class group_id;
    var items;
    output out=_means_ mean=;
run;
proc sort data=_means_ out=want;
    by descending items;
run;

This doesn't give the TOP N means, it gives you all means in descending order; but you can get the TOP N means easily from data set WANT in a DATA step.

[tmccombs27]

Thank you!

I applied this to my data set and I have what I wanted.  I used proc print to get my top 10 for display.  I have not used proc summary before, and I kept trying to use by where I should have used class.  I also would have never guessed that I could define a statistic within the OUT statement.

[PaigeMiller]

PROC SUMMARY is 99% the same as PROC MEANS. The default outputs are different, SUMMARY defaults to output to a SAS data set, MEAN defaults to output as text in HTML or LISTING.