In PROC FREQ, an option SPARSE can work your way; but you have to specify at least one time for those missed categories. For example:
Group /response /count
1 / medium/ 0
1 / low /0
The added lines won't impact on the frequency table but tell SAS indeed some more categories exist in the data.
jingju
比如
data;
input group response $ count;
cards;
1 high 100
1 medium 0
2 high 20
3 high 30
;
proc freq;
tables group* response/chisq sparse list out =a;
weight count /zeros;
run;