proc transpose倒是不用考虑name的数量，
只不过不能一步写成~

Hi Playmore,
I know the posting is back to last year. However I try to create a macro as an exercise.
If you have time, please go through it, and let me know any concerns. Thanks in advance.

data have;
input ByFactors $ Name $ MEAN STD;
datalines;
2009 A 23 41
2009 B 45 37
2010 A 27 53
2010 B 12 90
2012 C 23 35
2012 D 12 23
2011 A 23 45
;
run;

%MACRO createwant;
/*get the name with no duplicate*/
proc sort data=have out=have1 (keep=name) nodupkey;
by name;
run;

/*put all the name into a string*/
data _null_;
set have1 end=last;
length test $200;
retain test;
test=catx(' ',test,name);
if last then call symputx('mytest',test);
run;

/*Create the final dataset*/
data want;
%let i=1;
    merge
       %do %until(not %length(%scan(&mytest,&i)));
           have(where=(name="%scan(&mytest,&i)") rename=(mean=mean_%scan(&mytest,&i) std=std_%scan(&mytest,&i)))
           %let i=%eval(&i+1);
       %end;
    ;
    by byfactors;
    drop name;
run;

/*Delete the temp sataset*/
proc datasets library=work;
delete have1;
run;

quit;
%MEND createwant;

%createwant

多谢你的宏，作为范例写得很好啊

另外提一下，得到数据集的变量名称，用proc contents或者用SASHELP.VCOLUMN表都可以

还有当时我提这个问题是我需要处理这种不太符合数据库规范的表，
如果这个表包含如下变量就不用这么处理了
date variable statistic value
也就是在SAS中用proc means得到表
这种表处理起来就方便多了

还是谢谢你的回复，呵呵

最最mark！！！