calculating averages on floating period

[Ksharp]

So did you try my code. Just replace proc summary in my code.

data x(drop=month);
input DATE  : ddmmyy12.     MONTH  : monyy5.    VALUE ;
format date date9. month monyy5.;
cards;
01032011      MAR2011    8.50
02032011      MAR2011    8.50
02032011      MAR2011    5.50
02032011      MAR2011    8.50
14032011      MAR2011    7.50
24032011      MAR2011    9.00
24042011      APR2011     8.50
24042011      APR2011     8.50
26042011      APR2011     2.50
24042011      APR2011     3.50
16052011      MAY2011     2.50
24052011      MAY2011     13.50
25052011      MAY2011     4.50
26052011      MAY2011     3.50
27052011      MAY2011     2.50
25062011      JUN2011     10.50
26062011      JUN2011     13.50
27062011      JUN2011     12.50
28062011      JUN2011     13.50
;
run;
data x;
 set x;
 monyy=mdy(month(date),1,year(date));
 format monyy monyy5.;
run;




%macro cal(monyy=);
data xx;
 set x(where=(monyy between intnx('month',&monyy,-&bin+1,'b') and &monyy));
 month=put(&monyy,monyy5.);
run;
proc univariate data=xx noprint;
 class month;
 var value;
 output out=result mean=mean median=median  pctlpts = 10 to 100 by 10 pctlpre = pct_ ; 
run;
%mend cal;


%let bin = 3;

proc sort data=x(keep=monyy) out=temp nodupkey;by monyy;run;
data _null_;
 set temp;
 call execute('%cal(monyy='||strip(monyy)||')');
 if _n_ eq 1 then call execute('data final;set result;stop;run;');
 call execute('proc append base=final data=result force;run;');
run;

Ksharp

[Filipvdr]

Well i did not.. i went on with the code of Keith. If there is no answer within the code of Keith I will have to take yours

[Keith]

My method will not give you the percentiles you need, I stated the limitations of multilabel formats in my original answer.  Ksharp has provided an excellent solution which will work with UNIVARIATE.

Regards,

Keith

[Filipvdr]

Thanks both for your quick reply. I will have a look at Ksharp his code to understand it and implement it.