Date:         Fri, 12 Jan 2007 16:16:17 -0500
Reply-To:     Wensui Liu <liuwensui@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Wensui Liu <liuwensui@GMAIL.COM>
Subject:      Re: normality of residuals: opinions?
Comments: To: Kevin Roland Viel <kviel@emory.edu>
In-Reply-To:  <Pine.GSO.4.58.0701121546020.26687@leukothea>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

Kevin,

I think it is not sufficient to just look at resid.

If you use proc genmod instead of proc reg, you can use experimental
ASSESS statement to check the cusum of resid along your IV. If you see
an obvious trending of the cusum process, probably you need to
consider some kind of transformation for your IV.

Another way around is to use more flexible modeling methd such as proc
gam. Then you don't need to worry about transformation.

HTH.


On 1/12/07, Kevin Roland Viel <kviel@emory.edu> wrote:
> Sorry to burden the listserv like this, but I wanted to get a second
> opinion concerning the distribution of residuals from a REG procedure
> regression.  The question before me is whether I should log transform the
> IV.  The data follow my signature.
>
> Thank you,
>
> Kevin
>
> Kevin Viel
> PhD Candidate
> Department of Epidemiology
> Rollins School of Public Health
> Emory University
> Atlanta, GA 30322
>
> data resid ;
>   input e @@ ;
>   datalines ;
> -30.82552643 -21.3397105 -37.03212775 -6.420253898 -4.354539771
> -27.2464255 38.035531435 -23.69146352 -6.295435679 32.93315834
> -25.45056947 11.559333487 44.649008708 0.7401886498 -11.34197641
> 75.654991983 27.065480629 6.5301495758 25.116686186 -42.42766114
> -28.68553979 -4.458576944 9.4224713083 -8.277536206 -5.501828368
> -15.01747873 -9.267722824 8.2923232704 -48.70029662 -8.669815104
> -17.30691322 -14.45618494 -5.431568241 2.9701606043 -14.12689768
> 25.867878883 -5.549436378 -16.61588548 -5.140802417 14.190768964
> 26.19022189 -3.770819422 16.146057087 -17.41983058 18.742408056
> 24.064531123 8.3084245874 -13.57537732 -15.85003324 -2.502443873
> -5.359669937 43.545680751 8.3160052269 -12.14286283 25.592035526
> -15.55366141 -1.723769179 12.429297769 -31.86569333 72.064009603
> 6.1304383827 -18.81436701 1.7995950816 -11.30635246 34.688168224
> -10.86488365 -19.68057313 -10.07652613 -4.248687026 3.986568116
> 25.930058644 -4.489291724 -14.9272115 4.1168488701 -15.82953122
> -2.3813471 35.920661539 -14.98403741
> ;
> run ;
>
> proc univariate data = resid normal plot ;
>   var e ;
> run ;
>


--
WenSui Liu
A lousy statistician who happens to know a little programming
(http://spaces.msn.com/statcompute/blog)