Frank E Harrell Jr wrote:
>
> Whenever a multi-step strategy is used there should
> be a multi-step simulation to verify that statistical
> properties such as confidence coverage are preserved.
>
> Frank Harrell
>
I like that as a nice overall summary of how to approach stepwise
methods. It's clear, precise, short and leaves nothing unsaid.
I shall now cut-and-paste it (with attribution, of course) into my
class notes!
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
