If you think it worked before perhaps the data is different?  Maybe the value that looks like 1 is not exactly 1?   If is hard to tell what you want to happen.  But here is a way for you to examine the impact the tests you are doing.  Let's make a dummy dataset with some values of RULE and STATUS. data results; do rule='Rule 1','Rule 2','Other'; do status=1,2; output; end; end; run; Now let's create some binary 1/0 variables from the various conditions you are testing. data TEST; set RESULTS; r1= rule ne 'Rule 1'; r2= rule ne 'Rule 2'; r3= rule ne 'Rule 3'; s1= status ne 1; test1= r1 and s1; test2= r2 and s1; test3= r3 and s1; run; So the R1-R3 variables are the various RULE NE xxx conditions.  The S1 variable test if STATUS is no one. Then the TEST1-TEST3 show the impact of using AND with one of the R variables and the S variables. So TEST1 was what you original code was testing. Obs rule status r1 r2 r3 s1 test1 test2 test3 1 Rule 1 1 0 1 1 0 0 0 0 2 Rule 1 2 0 1 1 1 0 1 1 3 Rule 2 1 1 0 1 0 0 0 0 4 Rule 2 2 1 0 1 1 1 0 1 5 Other 1 1 1 1 0 0 0 0 6 Other 2 1 1 1 1 1 1 1 So which of these 6 observations did you actually want to select? ... View more