Date: Wed, 24 Dec 2008 08:56:02 -0500
Reply-To: Peter Flom <email@example.com>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Peter Flom <peterflomconsulting@MINDSPRING.COM>
Subject: PROC DISCRIM question - adjusting costs and priors
Content-Type: text/plain; charset=UTF-8
Happy holidays to all.
In discriminant analysis it is common to see discussion of a cost matrix ... that is, classify (say) a normal person as abnormal may be more costly than the reverse. In some programs (e.g. R) you can define this cost matrix directly. I do not see a way to do this in PROC DISCRIM. There are some connections between costs and priors, but it isn't clear (at least to me) exactly how to implement a cost matrix by changing priors.
I can just play around with different priors until the classification is as close to what I want as I can get, but I thought there might be some more systematic method, or perhaps I have overlooked an option on PROC DISCRIM
TIA as always
Peter L. Flom, PhD
www DOT peterflom DOT com