Date: Sat, 12 Aug 2006 06:14:57 -0400
Reply-To: Peter Flom <flom@NDRI.ORG>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Peter Flom <flom@NDRI.ORG>
Subject: Re: Hockey stick models
Content-Type: text/plain; charset=US-ASCII
Peter L. Flom, PhD
Assistant Director, Statistics and Data Analysis Core
Center for Drug Use and HIV Research
National Development and Research Institutes
71 W. 23rd St
New York, NY 10010
(212) 845-4485 (voice)
(917) 438-0894 (fax)
>>> David L Cassell <davidlcassell@MSN.COM> 08/12/06 2:00 AM >>> wrote
This is an important point I was making. So-called hockey-stick models
are way cruder than modern statisticians would like. They literally use
the concept I wrote out, with no quadratic terms. That's why I dropped
them out of your original code.
If you really want to handle some robustness and better curvature, then
I suggest cubic splines myself. And PROC TRANSREG does a fine job with
I agree, naturally.
But one good thing about hockey stick models is that they are easy to explain to non-statisticians.
And SOMETIMES they make sense. If you have a strong a priori reason for believing that
some regression will change at some particular point....
Like.....traffic accidents gotten into vs. age of getting a license. (and, of course, the bend in the stick doesn't have to happen at an AGE, it can happen at an event)
or dollars spent on travel vs. age, and retirement age
But generally I agree with David. They're simplistic