Date:         Tue, 12 Aug 2003 14:38:31 -0500
Reply-To:     Paul Thompson <paul@WUBIOS.WUSTL.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Paul Thompson <paul@WUBIOS.WUSTL.EDU>
Organization: Washington University in St. Louis
Subject:      Re: R-Square per variable
Content-Type: text/plain; charset=us-ascii; format=flowed

for a single variable, the partial R2 can be obtained by

SS-VARIABLE / SS-Total

Paul R Swank wrote:
> You can do that from the full model by solving for R sqaured change using
> the partial F test.
>
> F = {R2(change)/df(change)] / [(1 - R2(full) )/ df(full)]
>
> so
>
> R2(change) = F [(1 - R2(full) )/ df(full)] / df(change)
>
> Paul R. Swank, Ph.D.
> Professor, Developmental Pediatrics
> Medical School
> UT Health Science Center at Houston
>
> ----- Original Message -----
> From: "Jay Weedon" <jweedon@EARTHLINK.NET>
>
>
>>What you might find instructive is to look at the increase-in-R2
>>statistic associated with each variable conditioned on all other
>>variables already being in the equation. For instance, if you have
>>three predictors X1 X2 X3, you could run a model containing X1 & X2,
>>record the R2, and compare that with the R2 for the full model; this
>>will tell you how much explained variance is *added* by X3 on top of
>>that already explained by X1 & X2. You can do this for the other
>>variables as well.
>>
>>JW
>
>
>
> Jay,
>
>        Could this also be done using PROC VARCOMP, for instance?
>
> Thanks,
>
> Kevin
> ____________________________________
>
> Kevin Viel
> Department of Epidemiology
> Rollins School of Public Health
> Emory University
> Atlanta, GA 30329