Hi, No probs.  Let me explain the call execute first, this may clarify things.  Call execute takes a string and sends that string directly to the compiler rather than to the macro pre-processor.  The macro pre-processor is a bit like a find replace function where it expands all you code to full code before sending to the compiler, so it resolves macro variables, expands %if or %do loops etc.  What I am doing is creating the code I want to execute - and this can be any code - and utilizing the datastep which operates as a loop over each row to generate the code which gets sent to the compiler.  Here is an update following your request: data _null_;   set sashelp.vtable (where=(libname="WORK" and index(upcase(memname),"TEST")>0));   call execute('data work.exportit (keep=var1 var2 var5 var10); set work.'||strip(memname)||'; run;                        proc export data=work.exportit outfile="s:\temp\rob\'||strip(memname)||'.xls"; run;'); run; What this actually does is first creates a temporary dataset with all records from vtable where libname=WORK and TEST is in the dataset name.  Then for each of these rows creates a string - say we have two: x1_test and y1_test then there will be two strings sent to the compiler: "data work.exportit (keep=var1 var2 var5 var10); set work.x1_test; run; proc export data=work.exportit outfile="s:\temp\rob\x1_test.xls"; run;" "data work.exportit (keep=var1 var2 var5 var10); set work.y1_test; run; proc export data=work.exportit outfile="s:\temp\rob\y1_test.xls"; run;" If you check the log when you run something like the above, you will see what is generated under the 'NOTE: CALL EXECUTE generate line' ... View more

Hi, Have done the first three.  You can work out the fourth from the code, just need to mix up counts on the where - have to leave now. data have; attrib FIRM format=$1.  YEAR format=best.   NAME format=$200.   DUMMY    CHANGE format=best.; infile datalines delimiter=','; input firm $ year name $ dummy change; datalines; A,2004,Mr.X,0,0 A,2005,Mr.X,0,0 A,2006,Mr.Y,0,1 A,2007,Mr.Y,0,0 A,2008,Mr.Y,0,0 B,2004,Mr.S,0,0 B,2005,Mr.S,0,0 B,2006,Mr.T,1,1 B,2007,Mr.T,1,0 B,2008,Mr.T,1,0 C,2004,Mr.M,1,0 C,2005,Mr.M,1,0 C,2006,Mr.M,1,0 C,2007,Mr.N,1,1 C,2008,Mr.N,1,0 D,2004,Mr.R,1,0 D,2005,Mr.R,1,0 D,2006,Mr.V,0,1 D,2007,Mr.V,0,0 D,2008,Mr.V,0,0 ; run; proc sql;   create table subset1 as   select  *   from    have   group by firm   having   sum(dummy)=0;   create table subset2 as   select  *   from    have   group by firm   having   sum(dummy)=5; quit; data tmp;   set have;   length flag $4;   retain flag;   by firm;   if first.firm then flag='pre';   if change=1 then flag='on';   if change=0 and flag='on' then flag='post'; run; proc sql;   create table subset3 as   select  *   from    have   where   firm in (select distinct firm from (select * from tmp where flag="pre") group by firm having sum(dummy)=0)      and  firm in (select distinct firm from (select * from tmp where flag="post") group by firm having sum(dummy)>0); quit; ... View more