Date:  Tue, 6 Feb 2001 09:10:17 0600 
ReplyTo:  Jaclyn Whitehorn <jaclyn@BAMA.UA.EDU> 
Sender:  "SAS(r) Discussion" <SASL@LISTSERV.UGA.EDU> 
From:  Jaclyn Whitehorn <jaclyn@BAMA.UA.EDU> 
Subject:  Re: Computing the formula for complete and reduced regression
models in SAS 
InReplyTo:  <200102060600.AAA28638@bama.ua.edu> 
ContentType:  text/plain; charset="usascii"; format=flowed 

Jacinto:
I haven't used GLM for this, but you can use the TEST statement in PROC REG
(or the MTEST statement for multivariate regression). The documentation
should be sufficient to explain the syntax.
BTW, unless I'm having a worse brain day than I thought, you can think of
the F as a proportion reduction in MSE, so the formula is:
F = (MSE(reduced)  MSE(complete))/(MSE(complete))
where each MSE is the corresponding SSE divided by its degrees of freedom.
I THINK the degrees of freedom for the Fstatistic are the difference in
dfs (numerator )and the complete df (denominator).
And if I just screwed that up, I apologize to every one of my stats
professors ... :(
Jaclyn
>Date: Mon, 5 Feb 2001 19:18:55 0500
>From: Jacinto F Mendoza <leon_mendoza@JUNO.COM>
>Subject: Computing the formula for complete and reduced regression models
>in SAS
>
> I have been trying to figure out how to program in SAS the formula for
>the Ftest to test the complete and reduced regression models in SAS.
>The formula is easy to compute manually (after the 1 PROC GLM for the
>complete model and 1 PROC GLM for the reduced model). I think it is (if
>I can remember (I left my stat book at my job)):
>
>F= SSE(reduced m.)  SSE(complete m.)/df(complete m.)  df(reduced m.)
>
>
> SSE(complete m.)/n(df(complete m.)+1)
>
>Regardless of what the formula is, how do I get SAS to do the computation
>for me (instead of me doing it)? Any ideas?jfm
