Hi @ivanpersie,   It's good that you've provided sample data. Sample data in the form of a data step would have been even more convenient for people who want to test their suggested solutions: data have; input date :yymmdd10. VIC VNM VCB GAS; format date yymmdd10.; cards; 2014/12/1 28411.75 69096.69 20888.86 85000 2014/11/1 28762.52 72916.13 20888.86 105000 2014/10/1 28937.89 74305 19555.53 109000 2014/9/1 34199.33 77082.75 20740.71 124000 2014/8/1 29682.32 78124.06 19111.08 111000 2014/7/1 25357.74 70601 19259.23 111000 2014/6/1 26537.17 71758.44 18679.63 96500 2014/5/1 25554.31 80438.88 18550.8 99500 2014/4/1 28502.89 81596.25 20032.29 83000 ; You can use PROC TRANSPOSE to obtain the transposed dataset: proc transpose data=have out=want(rename=(_name_=company col1=price)); by date notsorted; var VIC--GAS; run; With the above dataset HAVE the BY statement could read by descending date; as well. In any case there must be no duplicate dates. A data step solution would be more robust to such duplicates and it would give you more control over length and label of variable COMPANY: data want; set have; length company $3; array p VIC--GAS; do i=1 to dim(p); company=vname(p[i]); price=p[i]; output; end; keep date company price; run;   ... View more

Sorry that program calculated 12 forthcoming months: This should calculate 12 previous: proc sql;   create table want as   select h1.ticker,          h1.date,          std(h2.return) as return_Std,          mean(h2.return) as return_mean,          min(h2.return)  as return_min,          max(h2.return)  as return_max,          n(h2.return)    as return_n   from   have h1 ,          have h2   where  h1.ticker = h2.ticker      and h1.date ge h2.date      and h1.date le h2.date+365.25 group by h1.ticker, h1.date; quit; ... View more

Check proc rank + ties=dense  ; ... View more