surely the forecasting application supports by-processing and classing?
You may find other forums more helpful.
Although it is a xxxxxxxx to manage them, it is easy to create hundreds of data sets[pre] data rest %gen( pattern= data.area###, from=1, upto= 1000 ) ;
set original.data_set ;
select (area) ;
%gen( pattern= %str( when ( ### ) output data.area### ; )
, from=1, upto= 1000 )
otherwise output rest ;
end;
run ;[/pre]
That will split the table original.data_set into up to 1001 datasets assuming that rows with "area" having value, say 23, should be written into table data.area23 .
Any rows with area outside the range 1-1000 will be written to table work.rest.
The code assumes you have a macro %GEN() available that works by replacing ### in a pattern with a number varying from &from to &upto. Something like:[pre]%macro gen( from=1, upto= 10, pattern= asdf###asdfg )/ des='pattern generator';
%local word pointer ;
%do pointer = &from %to &upTo ;
%sysfunc( tranwrd( %superq(pattern), ###, &pointer ))
%end ;
%mend gen ;[/pre]
but beware code offered over the internet you might want to test it on less than 100 datasets
PeterC