Date: Fri, 30 Sep 2005 10:46:45 -0400
Reply-To: Jerry Davis <jwd@GRIFFIN.UGA.EDU>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Jerry Davis <jwd@GRIFFIN.UGA.EDU>
Subject: Re: Bootstrapping a C statistic
Content-Type: text/plain; charset="US-ASCII"
David L Cassell wrote:
> ni14@MAIL.COM wrote:
> >I was reading this topic about Bootstrapping a C statistic from proc
> >logistic modeling. My question is, why does one wish to do bootstraping on
> >this? I don't see the meaning of this post because I don't know what it is
> >we are trying to see here.
> That may be fine for your purposes. And there may be no need to look at c
> in your situation. This was just a technical question.
But, asking for help with stepwise regression syntax is just a technical
> to consider that there is so much noise in the data that the estimate of c
> is totally
> unreliable. You really don't know whether your model is doing a great job
> or a
> truly lousy one. So using a bootstrap to estyimate the confidence interval
> is one
> way to assess how reliable your assessment statistics are.
It is a lousy way, IMHO. I don't see the utility of resampling from a
100 obs dataset, generating 1000 models and deriving statistical
inference. I just made up this example but bootstrapping seems to me to
be the epitome of beating data to death. Instead of measuring 50
subjects from a population, why not just measure one subject 50 times?
If the goal is more robust parameters estimates or diagnostics, then
collect enough data to get them.
Bootstrapping might be an interesting exercise, I just question its
usefulness. I'm glad that I rarely get asked about it. :-)
I'm not a theoretical guy so my comments are worth what you paid for
them. And for the record, I discourage the usage of stepwise regression
UGA, CAES, Griffin Campus