写一个循环就好了

DATA data;
INPUT index y x;
DATALINES;
1 100 400
1 120 380
1 351 684
2 454 243
2 334 343
2 343 343
;
RUN;
%macro a;
%do i=1 %to 2;
proc reg data=data(where=(index=&i));
model y=x;
run;
%end;
%mend;
%a;

1# a2zhou

"应该是使用循环控制语句来写，因为实际我的数据不止5组有1百多组，不可能一个一个搞，谢谢了！！！"

In this case you don't need any loops, instead using the "by statement in the proc reg" as in following example.


***use by statement in proc reg will have a regression eq for each by level;
***it will run much faster than macro approach as otherone proposed;
***generated data with 5 group;

data tmp;
  do i=1 to 1000;
     index=ceil(ranuni(12345)*5);   
     x=rannor(12345);
     y=1+2*x+rannor(12345);
     output;
  end;
drop i;
run;

proc sort data=tmp; by index;
run;

proc reg data=tmp;
by index;
model y=x;
run;
quit;

谢谢啦，我看了一下基础的sas的书，用by 就解决了，谢谢了。
这种base东西也只有看书才能找到了。