There are indeed situations like the concatenations you describe where the automatic retention of variables read with a SET statement can be somewhat perplexing. However, this has been standard SAS behavior for ages (or ever). It is described, for example, in the online documentation of the RETAIN statement, section "Redundancy": "variables that are read with a SET, MERGE, MODIFY or UPDATE statement" are among those whose "values are automatically retained from one iteration of the DATA step to the next." This is in fact a feature and it is probably used most in match-merging.   The important fact that @Astounding mentioned, "when you switch data sets, variables get reinitialized", seems to be harder to find in the documentation, but here it is: in the book Language Reference: Concepts. Quote: "SAS continues to read one observation at a time from the first data set until it finds an end-of-file indicator. The values of the variables in the program data vector are then set to missing, and SAS begins reading observations from the second data set ..."   The 2012 conference paper The Use and Abuse of the Program Data Vector (23 pages) takes a detailed look at the internal workings of the data step. In the section "Multiple data sets in the set statement" the author presents an example similar to yours. As far as I see, his argument is wrong where he tries to explain what happens with the first observation of the second dataset! He incorrectly states that a value "remains from the previous data step iteration." I think his own example data and code prove him wrong at this point. So, you see, you're not alone with your confusion.    In your second example VALSN is created with an assignment statement, hence it is not retained (like "all variables that are created with INPUT or assignment statements", as the online doc [see first link above] describes), unless you request this by using a RETAIN statement.   ... View more

Since I've never used DI studio, I can only guess. I take it that there must be some place in DI Studio where CASE expressions are valid, but the "New Value" field in the "Invalid Values" dialog box does not seem to be such a place. The code typed in there appears to be included into what looks like DATA step code. The first two error messages that you got (388-185 and 202-322) are exactly the same that I get if I try to assign a CASE expression to a variable in a DATA step. Indeed, "New Value" sounds to me more like a literal value or an expression using operators or functions. ... View more

How do you like this? /* Compute overall 25th percentile of score */ proc summary data=have; var score; output out=perc25 q1=q; quit; /* Write it into macro variable Q */ proc sql noprint; select q into :q from perc25; quit; %put Overall 25th percentile of score: &q; /* Create dataset with number of students per school */ data numstud; do nst=1 by 1 until(last.school_ID); set have(keep=school_ID); by school_ID; end; run; /* Add the desired new variables */ proc sql; create table want(drop=nst) as select *, (mean_score*nst-score)/(nst-1) as mrest, (mean_score*nst+&q)/(nst+1) as mwithq from have natural join numstud order by school_ID, student_ID; quit; Edit: Added optional ORDER BY clause. ... View more