About pedrogares

pedrogares · ‎12-07-2016

Dear all, Thanks for all your comments. I found that this temporary utility problem is mainly coming from the use of \view. I considered now all the observations and I did not have any problem about running proc corr statement (It ony took more time to do it). On the other hand, I used this 31 days assumption in order to take into account missing returns in some days. At the end, I am always imposing to get at least 750 trading days per stock but this way it is very easy to create the rolling windows. The only thing I do is to use a macro in order to compute the number of non-missing returns in these windows and impuse at least 750 days. Thanks a lot to all of you Best Pedro

pedrogares · ‎12-06-2016

Hi, I am trying to compute the BAB factor in the Betting against Beta paper. I am using overlapping 3 days log returns of the market returns and the stock returns. I am assuming that every month has 31 days and as they look for at least 3 years of data this is the way I arrive to 1116. I impose to get at least 750 trading days with non-missing returns as they require in the paper. Thanks to assuming this, I do not have problems when computing the rolling windows. The last step is to use the proc corr in order to arrive to the correlations. I am following what some of you recommend me I am geting rid of the \view and trying to run the proc corr for the essential elements I need x y PERMNO and N

pedrogares · ‎12-06-2016

Hi, I am trying to run PROC corr in order to obtain some rolling correlations based on 3 years. My main problem is that there is no enough internal space as the utility temporary file is really big ( I am dealing with more than 100M datapoints). Is there any way to avoid this utility temporary file to be saved on the SAS temporary files? I tried many things but neither of them worked. This is my code DATA rwin / view=rwin; array _X {1116} _temporary_ ; array _Y {1116} _temporary_ ; array _D {1116} _temporary_ ; set CORRELATIONS3; by PERMNO; retain N 0; N = ifn(first.PERMNO,1,N+1); I=mod(N-1,1116)+1; _X{I}=marketRET1; _Y{I}=stockRET1; _D{I}=day; if N>=1116 then do I= 1 to 1116; x=_X{I}; y=_Y{I}; d=_D{I}; output; end; run; proc corr data=rwin noprint NOPROB NOSIMPLE outp=Babcorr1(where=(_type_='CORR')); by permno N; var x; with y; quit;

Online Status	Offline
Date Last Visited	‎05-13-2020 09:33 PM

Re: Proc CORR and TEMPORARY UTILITY FILES

Re: Proc CORR and TEMPORARY UTILITY FILES

Proc CORR and TEMPORARY UTILITY FILES

Re: Proc CORR and TEMPORARY UTILITY FILES

Re: Proc CORR and TEMPORARY UTILITY FILES

Proc CORR and TEMPORARY UTILITY FILES