Here is the one that will do reasonable fast.

1) The idea is to have the table with all variables a1-a60 + b1-b40 with all values set to '0'.
2) check the observation with n >0
   if it is true, calculate the index for a and b to set the corresponding variables to 1, to repeat n times and to output it.
3) reset the updated variables in 2) to '0'




proc import datafile="C:\Downloads\100106170823e2156bd060f7b3.xls"
                    out=wksht
                    dbms=Excel replace;
run;

data needs;
array a(*) a1-a60;
array b(*) b1-b40;
retain a1-a60 b1-b40 (100*0);
set wksht;

if n>0 then do;
indexa=input(compress(upcase(x),'A'),best.);
indexb=input(compress(upcase(y),'B'),best.);
do i=1 to n;
       a(indexa) =1;
    b(indexb) =1;
    output;
end;
       a(indexa) =0;
    b(indexb) =0;
end;
keep a: b:;
run;

谢谢各位朋友的热心帮助！谢谢funnyxuke！bobguy！jingju11！
问题已解决！

版主，怎么把论坛币给热心人啊？