-----Original Message-----=0AFrom: SAS(r) Discussion [mailto:SAS-L@LISTSERV=
.UGA.EDU] On Behalf Of oslo=0ASent: Wednesday, November 11, 2009 11:45 PM=
=0ATo: SA
...@LISTSERV.UGA.EDU=0ASubject: Re: independent regression for fal=
se dicovary rate=0A=0ADear Arthur;=0A=0AThanks so much. I appreciate for yo=
ur help. I have checked. It worked perfectly. For example the pvalue from p=
roc corr is 0.1299 while pvalue from proc reg is 0.1298.=0A=0ARegards,=0A=
=0AOslo=0A=0A=0A=0A________________________________=0AFrom: Arthur Tabachne=
ck <art
...@NETSCAPE.NET>=0ATo: SA
...@LISTSERV.UGA.EDU; hok
...@YAHOO.COM=0AS=
ent: Wed, November 11, 2009 7:44:33 PM=0ASubject: Re: independent regressio=
n for false dicovary rate=0A=0AOslo,=0A=0ACan you get what you want by runn=
ing simple pearson correlation=0Acoefficients?=A0 If so, then you could jus=
t use something like:=0A=0Aods output FisherPearsonCorr=3Dcorr (keep=3Dvar =
pvalue);=0Aproc corr data=3Dsashelp.class fisher ( biasadj=3Dno )=0A=A0 noc=
orr noprint noprob nosimple;=0A=A0 var height--weight;=0A=A0 with age;=0Aru=
n;=0A=0Aand get what you want from the resulting work.corr file.=0AHTH,=0AA=
rt=0A--------=0AOn Wed, 11 Nov 2009 16:20:19 -0800, oslo <hok
...@YAHOO.COM>=
wrote:=0A=0A>=0A=0ASigurd;=0A=0AYes I am expecting linear dependence as we=
ll. However I have 36000 independent=0Avariable and only 350 observations. =
So I have no chance to run mutliple=0Aregression with stepwise or else. I c=
an run LASSO or LAR1. However I need=0Aselect independent variables=EF=BF=
=BDfor construct genetic algoritm=0Amodel=0A=0AThanks,=0A=0AOslo=0A=0A_____=
___________________________=0AFrom: Sigurd Hermansen <HERMA
...@WESTAT.COM>=
=0ATo: SA
...@LISTSERV.UGA.EDU=0ASent: Wed, November 11, 2009 3:23:06 PM=0AS=
ubject: Re: independent regression for false dicovary rate=0A=0AOslo:=0ALet=
me ask a na=EF=BF=BDve question if you will ...=0A=0AHow do you know that =
any and all linear combinations of 36000 variables are=0Aindependent? I've =
seen observed linear dependence with significant p-values=0Ain much smaller=
sets of randomly selected variables.=0AS=0A=0A-----Original Message-----=
=0AFrom: SAS(r) Discussion [mailto:SA
...@LISTSERV.UGA.EDU] On Behalf Of osl=
o=0ASent: Wednesday, November 11, 2009 1:05 PM=0ATo: SA
...@LISTSERV.UGA.EDU=
=0ASubject: independent regression for false dicovary rate=0A=0ADear SAS-L =
users;=0A=0AI have about 36000 independent variables (x1-x36000) and one de=
pendent=0Avariable (y). I need to run 36000 independent single regression m=
odels to=0Ause their p values in MULTTEST. So;=0A=0Aa) Is there any pratica=
l way to run 36000 independent single regression at=0Athe same time=0Ab) Is=
there any way to pick only p values from 36000 independent regression=0Amo=
del?=0A=0AHelps would be greatly appriciated,=0A=0AOslo=0A=0A=0A=0A
|
