| Date: | Sun, 19 Jan 1997 23:20:57 -0500 |
|---|
| Reply-To: | tom <tjb@ACPUB.DUKE.EDU> |
|---|
| Sender: | "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU> |
|---|
| From: | tom <tjb@ACPUB.DUKE.EDU> |
|---|
| Organization: | Duke University, Durham, NC, USA |
|---|
| Subject: | Re: [q] crossvalidation on logistic regression |
|---|
In article <5bondg$f24$1@mark.ucdavis.edu>, szbchan@chip.ucdavis.edu
(Benjamin Chan) wrote:
> For observations in your validation sample, set a weight variable to 0.
> For observations in your testing sample, set the weight to 1. Concatenate
> the two datasets and run PROC LOGISTIC with the WEIGHT statement. The
> model fit will exclude validation records (since these are weight by 0).
> However, the OUTPUT statement calculates predicted values, xbetas, etc, on
> the entire dataset.
This worked. Cool. Thanks a lot.
Tom
|
