OK.just add one more clause.   proc sql; create table have as select *, (select std(return) from have where permno=a.permno and date between intnx('month',a.date,-60) and a.date) as rolling_std, (select count(*) from have where permno=a.permno and date between intnx('month',a.date,-60) and a.date) as n_rolling_std from have as a; quit; ... View more

@cm3 wrote: Hi    I’ve data like below   accno.  Var1 var2 var3  111.       Y.     Y.     N 112.       Y.     Y.     Y 113.       N.    Y.      Y 114.       Y.     N      N 115.       Y.     Y.      Y 116.       Y.     Y.     Y 117.       Y.     Y.     N 118.       Y.     Y.     Y 119.       Y.     Y.     Y 120.      N.     Y.     Y   i need output as below by considering only Y values for Var1,... tot.   %var1.   %var2. %var3 10.     80.          90.       70     please advise. Thank you   Are the values "Y"  or "Y."  That is two different values. Similar with "N." and "N". So do you want percent of "Y" or "Y."?   I  you use 1/0 numeric coding instead of Y/N character coding (or read external data of Y/N into 1/0) this becomes a trivial exercise. The SUM of 1/0 codes values becomes the number of 1 (or Yes or Y or True or what have you responses) and the mean is a percentage. ... View more

Please explain what you are trying to test for.  Since a variable can have a different value on each observation in a dataset what do you mean by a blank variable?  Do you want to know if it ever is blank? Or if it always is blank? What do you mean by blank?  Are you only testing character variables?  What about numeric variables?  Do you want to test for missing values?  What about special missing values?   If your variables do not have too many levels then a quick way to test if there are any missing (and any non-missing) values is to use the NLEVELS option on PROC FREQ.     proc freq data=have nlevels ; tables varname / noprint; run; ... View more

data have; input mr_name $19.; cards; cxhimgqabout lolfvchickbznicefx howmefvtoldbqcv toldyounothing ; run; data bigram; input bigram $; cards; bq bz cf cj cv cx fq fv fx fz gq ; run; data want(drop=bigram); if _n_=1 then do; if 0 then set bigram; dcl hash H (dataset:'bigram') ; h.definekey ("bigram") ; h.definedata ("bigram") ; h.definedone () ; dcl hiter hh('h'); end; set have; do while(hh.next()=0); count=sum(count,count(mr_name,strip(bigram))); end; run; ... View more

Try this... data sample; infile datalines dlm="|"; input Case:$30. vars: $30. Y112: 8. Y113:8. Y114:8. ; datalines; Basic|Temperature|20|23|25 Basic|Speed|130|120|80 Serious|Temperature|120|123|125 Serious|Speed|230|220|280 ; run; proc sort data=sample; by case ; run; data all; merge sample(where=(vars='Speed') rename=(Y112=Speed1 Y113=Speed2 Y114=Speed3)) sample(where=(vars^='Speed') rename=(Y112=Temp1 Y113=Temp2 Y114=Temp3 )); by case; drop vars; run; ... View more

Proc report data = final nowd; columns A B C ("Min %left(&MinWeight)" Name ("Activity" Amount Percent) ("May" Result)) ; run; ... View more