I hope the following works for your requirement.    data have; input company date:yymmdd10. date1:yymmdd10. var1; format date date1 yymmdd10.; datalines; 1 19960630 19960624 5 1 19960630 19960625 6 1 19960630 19960626 7 1 19960630 19960627 2 1 19960630 19960628 0 1 19960630 19960701 9 1 19960630 19960702 6 1 19960630 19960703 1 1 19960630 19960705 5 1 19960630 19960708 8 2 20041231 20041227 2 2 20041231 20041228 6 2 20041231 20041229 9 2 20041231 20041230 5 2 20041231 20041231 3 2 20041231 20050103 8 2 20041231 20050104 7 2 20041231 20050105 6 2 20041231 20050106 4 2 20041231 20050107 0 ;    data want;    if _N_ = 1 then do;    if 0 then set have;      declare hash h(dataset: "have(keep=company date1 var1)",multidata:'yes');       h.defineKey('date1');       h.defineData('var1','date1');       h.defineDone();    end;    do until(last.company);        set have(keep=company date);        by company;          if first.company then             do;                    call missing(__flag);                    do _temp=date by -1 while(1);                   rc=h.find(key:_temp);                   if rc=0 then                   do;                           _flag+1;                             if _temp=date then __flag=1;                             output;                           call missing(var1,date1);                   end;                        if __flag=1 and  _flag=3 then                    do;                             _flag=0;                               leave;                         end;                        else if missing(__flag) and _flag=2 then                         do;                           _flag=0;                              leave;                         end;                                           end;                   do _temp=date+1 by 1 while(1);                   rc=h.find(key:_temp);                   if rc=0 then                   do;                           _flag+1;                      output;                           call missing(var1,date1);                   end;                        if _flag=2 then                    do;                             _flag=0;                               leave;                         end;                   end;               end;               else leave;      end;      drop _:  rc;      run;   Regards, Naveen Srinivasan ... View more

Hi , Sorry for jumping in a bit too late. I have a feeling, you could make the solution less tedious and more effective by using Hashes or formats( as @Reeza )suggested. I hope your solution is working, nevertheless the logic can be hard on the macro processor and the storage of large sets of variable lists in macro vars. Precisely, that invokes heavy macro processor work even before the datastep compiler is involved.    I haven't been active here lately however if I find time after my Final exams, I will try out the Hash solution and post it here. Anyways, if i missed out a point or two to read between the lines, kindly ignore my post. My apologies.   Regards, Naveen Srinivasan ... View more

OK, I trust you attempted all the well laid out guidelines of Reeza. Here is the Hash solution. Depending on the data types of variables in your HAVE dataset, you should be able to make slight adjustments because my testing is limited to using only your sample.   data have; input Id $ gender_cd $ martial_status $ currency_cd $ ; datalines; 1 M 0 10 2 V 1 20 3 G 2 30 4 M 1 20 ; run;         data master; length column $14.; input column $ text $ m_cd $; datalines; Gender_cd M 1 Gender_cd V 2 Gender_cd G 4 Martial_status 0 bachlor Martial_status 1 single martial_status 2 married currency_cd 10 USD Currency_CD 20 INR Currecny_CD 30 AUD ; run;   data want;      if _N_ = 1 then            do;                 if 0 then do;                 set have;                 set master(keep=text m_cd);                 end;                 declare hash myhash(dataset:'master(keep=text m_cd)',multidata:'yes' );                 rc = myhash.defineKey('text');                 rc = myhash.defineData('m_cd');                 rc = myhash.defineDone( );            end;      set have;      array list(*) gender_cd--currency_cd;     do _n_=1 to dim(list);            rc_temp=list(_n_);             if myhash.find(key:rc_temp)=0 then                 list(_n_)=m_cd;            else call missing(m_cd);      end;      drop rc: m_cd text; run;     Regards, Naveen Srinivasan ... View more

Another fun way: /*Assuming the number of records in each group is divisible by 3 as it is in your sample to conveniently make it sets of three*/ data want;      do until(last.id);            do _n_=1 by 1 until(mod(_n_,3)=0);                 set dates;                 by id notsorted;                 array temp{3} _temporary_;                 temp(_n_)=point;                   if last.id then                      do;                            sum=sum(of temp(*));                            output;                      end;            end;      end; run;   Regards, Naveen Srinivasan ... View more

Ok, if you are not keen on explaining the real situation, here is the answer for your question. However, folks would be more interested to help and solve real problems.   　     data have; input Var1 Var2; datalines; 4 10 7 20 9 30 3 40 2 50 5 60 ;   data want;      set have;      call symputx('macrovar',Var1);      Var3 = Var2 + resolve('&macrovar') ; /*equivalent of Var3 = Var2 + Var1*/ run;   Regards, Naveen Srinivasan ... View more