Hi. I will try to answer to some/all of your questions, but first there is no PROC Merge, rather a DATA step. MERGE is the statement. When you run the code data state.a; merge state.a(in=tag) state.b by id; if tag=1; run; two things will happen: 1. you only keep the observations that are in A (I assume that is what you want) and 2. for any observation in both A & B, if your two data sets have same variables, then state.b will overwrite the values in state.a For example, state.a has ID Age Height 1     23     5.3 2     .     5.3 And state.b has ID Age Height 1     23     5.3 2     34     5.3 then when you merge these two data sets, the final data set state.a will have ID Age Height 1     23     5.3 2     34     5.3 Which may or may not be right. I would suggest renaming all of your common variables to make sure that you do get the results you want and not overwrite values. Working in WORK library or not has no impact on your data step. Also, I assume that you are merging 1-to-1 or 1-to-many, right? Anyway, I hope this helps a bit? Good luck and let me know if you have more questions. Anca. ... View more

Tal, What LinLin has written is correct (and everyone else). If you want to create a table from SQL you need to tell SAs to save sum(x) as sum_X and sum(y) as sum_Y I am just making sure you noticed 'extra' ASes. Best of luck. Anca. ... View more

So, I feel like I got it to work. *Create a temp data that I will use to insert; data t1;     do i = 1 to 10;         diagnosis1 = i;         diagnosis2 = i-2;         diagnosis3=5;     output;     end; run; *the data I want to insert into; data t2;     do i = 1 to 20;         diagnosis = i;     output;     end; run; options mprint mlogic symbolgen; %macro loop(file, count) ;                 %do i = 1 %to &count. ;                 proc sql ;                         insert into work.t2                         select  i as personid, diagnosis&i. as diagnosis                         from &file                 where  diagnosis&i. <> . ;                quit ;                 %end ; %mend loop ; %loop(t1, 3) ; So, thy this and see if it works. I feel like I use exactly the same code.; :smileyconfused: ... View more

hmm, something like this may work: proc sort data = bob;by subjid lbtestcd lbdy;run; data bob_1;     set bob;     by subjid lbtestcd lbdy;         prev_e = lag(flag1);         prev_visit = lag(lbdy);         if first.subjid then do;                 prev_e = .;prev_visit = .;         end; yes_prior = (prev_e = flag1);*flag that there is a prior 'e'; consecutive = ( lbdy - prev_visit = 1);*flag that there are consecutive visits; what_you_want = (yes_prior = 1 & consecutive = 1);*the flag you are looking for, I think; run; Best of luck! Anca. ... View more

Hi. You can use the DAY function from SAS. day_part = day(time_variable); You could attach a sample data (say 10 observations max, and make sure you de-identify the data) Good luck. ... View more