In your code, macro variable has been reassigned 3 times, in the end, it took the last value. You probaly will need; data _null_;   length top3 $50;   do _n_=1 to 3;   set customer_sum ;   top3=catx(', ', top3, Customer_ID);   end;   call symputx('TOP3',top3); run; or: proc sql; select Customer_ID into :top3 separated by ' ,' from customer_sum(obs=3); quit; Regards, Haikuo Edit : to make your downstream code working: proc print data=orion.customer_dim; var Customer_ID Customer_Name Customer_Type; where Customer_ID in (&TOP3); run; ... View more

Quite a mystery to me.  I know this is silly, as I can't copy OP's environment,  but I can't help trying it out and it is not happening to me. 255  data have; 256  array v(37) co Year Mon ped ccm gas gash gast gasv gasdd oil oilh oilt oilv oildd cnd cndv 256! cnddd 257     othinc othincdd totrev totdd loe gen int tax dda dde uop aro 258     totexp netinc mcfe taxcon taxprp taxsev taxest ; 259  do j=1 to 10; 260  do i=1 to 37; 261      v(i)=i; 262      end; 263      output; 264   end; 265   run; NOTE: The data set WORK.HAVE has 10 observations and 39 variables. NOTE: DATA statement used (Total process time):       real time           0.01 seconds       cpu time            0.01 seconds 266 267   data want; 268   set have; 269  keep co Year Mon ped ccm gas gash gast gasv gasdd oil oilh oilt oilv oildd cnd cndv cnddd 270     othinc othincdd totrev totdd loe gen int tax dda dde uop aro 271     totexp netinc mcfe taxcon taxprp taxsev taxest; 272     run; NOTE: There were 10 observations read from the data set WORK.HAVE. NOTE: The data set WORK.WANT has 10 observations and 37 variables. NOTE: DATA statement used (Total process time):       real time           0.01 seconds       cpu time            0.01 seconds Every thing seems normal to me. ... View more

I don't know about nested case-when. The following code seems to be equvalent to your datastep code. proc sql; create table as select *,        case             when city_code = 100 and Rank_100 in (1,2,3)             then '1_3'             when  city_code = 100 and Rank_100 in (4,5,6)             then '4_6'             when city_code = 100 and Rank_100 in (7,8,9)             then '7_9'             when city_code = 100 and Rank_100 not in (1,2,3,4,5,6,7,8,9)             then  'NO_Rank'             when city_code = 101 and Rank_100 in (1,2,3)             then '1_3'             when city_code = 101 and Rank_100 in (4,5,6,7,8)             then '4_8'             when city_code = 101 and Rank_100 in (13,14,15)             then '13_15'             when city_code = 101 and Rank_100 not in (1,2,3,4,5,6,7,8,13,14,15)             then 'NO_Rank'             when city_code = 102 and Rank_100 in (20,21,22)             then '20_22'             when city_code = 102 and Rank_100 in (23,24,25,26)             then '23_26'             when city_code = 102 and Rank_100 in (27,28,29,30)             then '27_30'             when city_code = 102 and Rank_100 not in (20,21,22,23,24,25,26,27,28,29,30)             then 'NO_Rank'             else 'Others'     end as     Rank_rng length=7     from population_a     ;     quit; Update: a quick search showing nested case-when is actually supported: http://www.nesug.org/Proceedings/nesug97/posters/zhang.pdf Good Luck, Haikuo ... View more

There you go! ... View more

Well, then: proc sql; update table yourtablename set desc=            case                 when cd is null   then 'n/a'                  when strip(cd)='1'   then 'minor'   when strip(cd)='2'   then 'moderate'   when strip(cd)='3'   then 'major'   when strip(cd)='4'   then 'extreme'   else 'others'   end;   update table want    set cd='0' where desc='n/a'; quit; ... View more

Try this one: data h1; input id $ date : date9. product :$ amount; cards; 1 01jan2012 box 20 1 03jan2012 ball 5 1 04jan2012 chair 10 2 02feb2012 camera 100 2 03feb2012 cup 4 ; data h2; input id $ date :date9. product :$ amount; cards; 1 03jan2012 refund -5 2 03feb2012 refund -4 ; data want; merge h1 (in=h1) h2(in=h2); by id date; if (h1 and not h2); format date date9.; run; proc print;run; Regards, Haikuo Edit: if you want to play with hash, here is one of them: data want;   if _n_=1 then      do; set h2(obs=1); dcl hash h(dataset: 'h2', multidata:'yes'); h.definekey('id','date'); h.definedone();    end; format date date9.;    set h1;    _n_=h.find();    if _n_ ne 0;    run; ... View more

You need to let us know the variable type of your variables: cd and desc. is 'cd' numeric? You find the info by running: 1. proc sql; describe table yourtablename; quit; or 2. proc contents data=your tablename; run; After knowing the types of your variables, you will be able to make changes. Regards, Haikuo ... View more

I like this one, avoiding hard-coding is better. ... View more

My bad, I have checked your code again, there was an error: data _null_; set geog.usa; length _list $ 4000; array _x{*} _numeric_; it should be: data _null_; set temp; length _list $ 4000; array _x{*} _numeric_; That is the whole purpose of temp. ... View more

If you criteria is 69711, then you should not use 'eq', which only took out those variablers with exact 69711 missing obs. Instead, you should use 'ge' or '>=', if _x{i} >= 69711 then _list=catx(' ',_list,vname(_x{i})); Also, take out the following statements, since you are not using the total number of obs as criteria, that will save you some time: select count(*) into : nobs from GEOG.USA; Haikuo ... View more