Help me understand where my logic differs from yours:   proc sort data = have; by id group descending var; run; The makes it so that the data properly organized by id and group.  "descending var" means that the first value of var for the id+group combination is going to be the largest in the group.     data want; set have; retain new_var 0; new_var = max(var, new_var); run; This data step does both of your rules at once.  Since the largest value of the group is first, the max function will assign all of the same id+group combinations the same value.  Since max is a monotonic function, new_var will only increase when the next group has a larger value.   Perhaps you want new_var to reset for each id, then we need to add some extra code: data want; set have; by id; retain new_var; new_var = ifn(first.id, var, max(var, new_var)); run; Now the data step will set the value of new_var to var at the first row of an id, and then use the larger of var and new_var thereafter. ... View more

I think what Ipy0521 has a typo there. For both rules, it should be new_var instead of var. And specifically for 2, I think it should be "new_var should be the same value for the same id+group pair, e.g. both rows with id = AAA and group=group2 have new_val=0.2" ... View more

The lines:   retain new_var 0; new_var = max(var, new_var); first create new_var, setting its value to 0 and it tells SAS that we want to keep the value of new_var when we move to the next row. Then the second line tells SAS to compare new_var's value with the current row's value for var and assign new_var the larger of the two. The result of my data step produces an identical data sets.  Validation code here:   Data Have; INPUT id $ group $ var ; Cards; AAA group1 0.1 AAA group2 0.2 AAA group2 0.2 AAA group3 0.3 AAA group3 0.2 AAA group3 0.3 AAA group4 0.4 AAA group4 0.4 AAA group5 0.3 BBB group1 0.4 BBB group2 0.3 BBB group2 0.2 BBB group3 0.3 BBB group3 0.3 BBB group3 0.2 BBB group4 0.3 BBB group4 0.2 BBB group5 0.3 ; RUN; proc sort data = have; by id group descending var; run; data want; set have; retain new_var 0; new_var = max(var, new_var); run; Data WANT2; INPUT id $ group $ var new_var; Cards; AAA group1 0.1 0.1 AAA group2 0.2 0.2 AAA group2 0.2 0.2 AAA group3 0.3 0.3 AAA group3 0.2 0.3 AAA group3 0.3 0.3 AAA group4 0.4 0.4 AAA group4 0.4 0.4 AAA group5 0.3 0.4 BBB group1 0.4 0.4 BBB group2 0.3 0.4 BBB group2 0.2 0.4 BBB group3 0.3 0.4 BBB group3 0.3 0.4 BBB group3 0.2 0.4 BBB group4 0.3 0.4 BBB group4 0.2 0.4 BBB group5 0.3 0.4 ; RUN; proc sort data = WANT2; by id group descending var; run; proc compare base = want comp = want2; run; ... View more

Here is a shot in the dark for you.  This appears to solve your problem: proc sort data = have; by id group descending var; run; data want; set have; retain new_var 0; new_var = max(var, new_var); run; ... View more

Just like Reeza said, it appears that you have duplicate values for the joined variable in at least one of the tables that you are joining. Here is some code to show this happening: data work.have1; col1 = 1; col2 = 1; output; col1 = 2; col2 = 1; output; col1 = 3; col2 = 1; output; run; data work.have2; col1 = 1; col3 = 2; output; col1 = 2; col3 = 2; output; col1 = 4; col3 = 2; output; run; proc sql; create table work.want as select * , b.col3 from work.have1 as a left join work.have2 as b on a.col1 = b.col3 ; quit; Even though we are joining two tables with 3 rows each, the table created has 5 rows. ... View more

Just an observation, the folders are different in the two screenshots.  The one with data is 949968 and the one shown in the EM screenshot is 1000495. ... View more