Whilst there may be some technical part of proc report that can do this for you, I always do my data processing before the reporting stage for a number of reasons. This will for instance get results for your test data you posted (good to post it as a datastep for future reference):
data have;
input name $ trials benchpress legpress;
datalines;
gary 2 170 100
gary 3 150 90
gary 4 160 89
peter 1 180 90
peter 7 120 80
tom 3 100 50
tom 4 110 60
tom 5 80 40
;
run;
proc sql;
create table WANT as
select NAME,
"benchpress" as TYPE,
mean(BENCHPRESS) as MEAN
from HAVE
group by NAME
union all
select NAME,
"legpress" as TYPE,
mean(LEGPRESS) as MEAN
from HAVE
group by NAME
order by NAME,TYPE;
quit;
However I suspect you have far more categories than that, so I would suggest going for a normalised data structure as this will make your coding so much easier.
data have (keep=name type val);
input name $ trials benchpress legpress;
do type="benchpress","legpress";
val=ifn(type="benchpress",benchpress,legpress);
output;
end;
datalines;
gary 2 170 100
gary 3 150 90
gary 4 160 89
peter 1 180 90
peter 7 120 80
tom 3 100 50
tom 4 110 60
tom 5 80 40
;
run;
proc sort data=have;
by name type;
run;
proc means data=have noprint;
by name type;
var val;
output out=want mean=mean;
run;
I have changed the test data to show a normalised structure, what it does is enable by group processing, which saves your program.
... View more
