Jerry Dallal <[EMAIL PROTECTED]> wrote in message
news:<[EMAIL PROTECTED]>...
> Rich Ulrich wrote:
> >
> > On Mon, 03 Dec 2001 20:57:33 GMT, Jerry Dallal
> > <[EMAIL PROTECTED]> wrote:
>
> > > This is a bit different, though. The criticisms of stepwise
> > > selection are directed toward the attempt to assess the contribution
> > > of individual predictors. The predictions themselves should be
> >
> > I don't see much difference. "Identifying predictors" by regression
> > analyses -- what is that advice supposed to mean? The criticisms
> > of stepwise selection say that it gives you the wrong variables, not
> > 'merely' (as if that were trivial) the wrong weights.
> >
> > Am I missing something? (I am not totally against stepwise;
> > just, mostly.)
> >
>
> There are two issues: determining the right set of variables and
> predicting the response. Stepwise can be deadly for the former
> without, I believe, being too bad for the latter. I'm willing to
> recant if someone with authority claims otherwise.
I don't claim to be an authority (and I don't much like arguments
from authority). But I'm going to put in my two cents worth.
I think it's just as dangerous with prediction as it is with
variable selection. The problem is this: within-sample, a good
fit is relatively meaningless, as with a large random set of
predictors we can get a good fit. For the purpose of prediction,
it's almost always out-of-sample predictive ability that's important.
And for that, stepwise can be bad, in exactly the same way it
can be bad for variable selection - it can leave you with
variables that are useless out-of-sample because they are just
fitting your noise term in-sample.
Glen
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
