Two ways to do this, with or without SAS/ETS : data have; format date yymmdd10.; input STOCK $ DATE :mmddyy10. STOCKRETURN MARKETRETURN; datalines; AAA 6/15/2012  0.19 0.29 AAA 6/16/2012 0.19 0.29 AAA 6/17/2012 0.2 0.3 AAA 6/18/2012 0.2 0.3 AAA 6/19/2012 0.21 0.31 AAA 6/20/2012 0.21 0.31 AAA 6/21/2012 0.2 0.3 AAA 6/22/2012 0.19 0.29 AAA 6/23/2012 0.19 0.29 AAA 6/24/2012 0.19 0.29 AAA 6/25/2012 0.19 0.29 AAA 6/26/2012 0.2 0.3 AAA 6/27/2012 0.2 0.3 AAA 6/28/2012 0.2 0.3 AAA 6/29/2012 0.2 0.3 ABC 6/15/2012 0.19 0.29 ABC 6/16/2012 0.21 0.31 ABC 6/17/2012 0.21 0.31 ABC 6/18/2012 0.19 0.29 ABC 6/19/2012 0.2 0.3 ABC 6/20/2012 0.21 0.31 ABC 6/21/2012 0.21 0.31 ABC 6/22/2012 0.21 0.31 ABC 6/23/2012 0.24 0.34 ABC 6/24/2012 0.25 0.35 ABC 6/25/2012 0.25 0.35 ABC 6/26/2012 0.25 0.35 ABC 6/27/2012 0.24 0.34 ABC 6/28/2012 0.24 0.34 ABC 6/29/2012 0.24 0.34 ; proc sql; create table have0 as select a.stock, a.date, a.stockreturn, intck("DAY", a.date, b.date) as lag,      b.marketReturn as mr from have as a inner join have as b      on a.stock=b.stock and intck("DAY", a.date, b.date) between -2 and 2 order by a.stock, a.date, lag; proc transpose data=have0      out=want(drop=_name_) prefix=mr; by stock date stockreturn; id lag; var mr; run; /* Or, if you have SAS/ETS, use proc expand */ proc expand data=have out=want; by stock; id date; convert marketreturn=mr_2 / transform=(lag 2); convert marketreturn=mr_1 / transform=(lag 1); convert marketreturn=mr0; convert marketreturn=mr1 / transform=(lead 1); convert marketreturn=mr2 / transform=(lead 2); run; /* Do the regressions, requesting the Durbin-Watson test for autocorrelation */ proc reg data=want; by stock; model stockreturn = mr_2 mr_1 mr0 mr1 mr2 / dwProb; run; PG ... View more