ｐｒｏｃ　ｅｘｐａｎｄ？？

%macro rollreg(datain=);

  %do firstmonth = 200101 %to 2004012;
    %let lastmonth = %eval( &firstmonth + 12);
    data temp; set &datain;
      where &firstmonth<= month <= &lastmonth;
    proc reg data=temp;
       * add your model statement here;
    run;
  %end;

%mend rollreg(datain=);

%rollreg(data=YOURDATANAME); 参考人大经济论坛~

这就是传说中的rolling regress吗？

Reshape your data then use reg to estimate all of them. It is faster, simpler in logic. Here is an example.

data testdata;
   do date=today()-29 to today();
      x1=rannor(123);x2=rannor(123);err=rannor(123);
          y=1+2*x1+0.5*x2+err;
          output;
        end;
format date yymmdd10.;
run;

data rollddata;
   do p1=1 to n-11;
      do p2=p1 to p1+11;
             set testdata point=p2 nobs=n;
                 output;
          end;
        end;
        stop;
run;

proc reg data=rollddata outest=parm noprint;
by p1;
model y=x1 x2;
run;
quit;

proc print data=parm;
run;

proc means data=parm n mean std;
var Intercept       x1         x2
;
run;