To help with your Googling, this technique (removing the extreme observations) is known as "trimming".  You can use PROC UNIVARIATE to get the trimmed means. You can read about some of the statistical shortcomings that are associated with trimming (or "Winsorizing") data,   If you are trying to perform a robust ANOVA, the best way is to use a statistical procedure that is not sensitive to outliers. There is an example in the SAS documentation that shows how to use PROC ROBUSTREG to perform a robust ANOVA. By using PROC ROBUSTREG, you will not throw away good data unnecessarily, which is what happens when you automatically discar the minimum and maximum values in each group.  ... View more

I nominated data step, since you only want proc print. data subsys;   infile datalines dlm=' ';   input SUBSYS $ CAP ALLOC AVAL PERCENTUSED; return; datalines; 2107#1          67.94     49.26     18.68       72.5  2107#2          67.94     50.93     17.01       75.0  2107#3          69.39     39.54     29.85       57.0  2107#4          67.24     24.94     42.30       37.1  ; run; data want(drop=cap_sum alloc_sum aval_sum) ; set subsys end=last; cap_sum+cap;alloc_sum+alloc; aval_sum+aval; output; if last then do;                 subsys='Total';                 cap=cap_sum;                 alloc=alloc_sum;                 aval=aval_sum;                 percentused=100*alloc_sum/cap_sum ;                 output;                end; run; proc format; picture fmt    low - high ='09.9%'; run; proc print data=want noobs; format percentused fmt.; run; Ksharp ... View more

Works great and a great help - now I have to implement across hundreds of emails per day...thanks ... View more