Date:         Mon, 16 Jul 2007 14:29:59 -0700
Reply-To:     jn mao <jn_mao@YAHOO.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         jn mao <jn_mao@YAHOO.COM>
Subject:      Re: question about stepwise in poisson regression
Comments: To: Peter Flom <peterflomconsulting@mindspring.com>
In-Reply-To:  <6128882.1184619938584.JavaMail.root@mswamui-billy.atl.sa.earthlink.net>
Content-Type: text/plain; charset=iso-8859-1

Thanks Peter,

  I have 200 data and 11 variables, including demographic variables and clinical variables. The 4 of the 11 variables are continuous, and the others are coded as 0 or 1. I'm trying to find the final model with all significant predictors. Now, when I do the manual stepwise, like move some nonsignificant ones, then add back some of them, back and forth, the signficant vars in the model always changed, I'm not sure which variables should be added in the significant model. Please give me some advice. Thanks.

  Jane




Peter Flom <peterflomconsulting@MINDSPRING.COM> wrote:
  >Hello SAS-Ls,
>
> I have 11 covariates in a possion model. Could someone give me some advise on STEPWISE selection? Seems SPTEWISE selection can't be used in poisson model. How should I find the final model? Thanks much.
>
> Jane
>
>

First advice: Don't use stepwise.
David Cassell and I are working on a paper (accepted at NESUG, already given at NYASUG) entitled: Stopping Stepwise.

You can read it here:

http://www.sascommunity.org/wiki/Image:NYASUG-2007-June-Flom-Stopping_stepwise.pdf

Now, if you tell us more about what your data are like, what N is, what your DV is, what you're trying to find out, and so on, then we can provide advice. Please write to SAS-L, not just to me.

Peter



---------------------------------
Park yourself in front of a world of choices in alternative vehicles.
Visit the Yahoo! Auto Green Center.