Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: Notesfiles $Revision: 1.7.0.10 $; site uiucdcsp
Path: utzoo!watmath!clyde!cbosgd!ihnp4!inuxc!pur-ee!uiucdcs!uiucdcsp!leimkuhl
From: leimkuhl@uiucdcsp.CS.UIUC.EDU
Newsgroups: net.math
Subject: Re: simplex algorithsm for curve fittin
Message-ID: <9600041@uiucdcsp>
Date: Sun, 16-Feb-86 14:44:00 EST
Article-I.D.: uiucdcsp.9600041
Posted: Sun Feb 16 14:44:00 1986
Date-Received: Tue, 18-Feb-86 03:50:07 EST
References: <1217@princeton.UUCP>
Lines: 11
Nf-ID: #R:princeton.UUCP:1217:uiucdcsp:9600041:000:440
Nf-From: uiucdcsp.CS.UIUC.EDU!leimkuhl    Feb 16 13:44:00 1986




One of my roommates is a biophysicist.  He's been using that program
to study models lately.  The instability for some problems is the result
of local minima on the surface associated with the curve.  You can get
reasonable confidence by "shaking" the surface after simplex settles down
(i.e. perturb the data and restart).  And of course you can always do a
goodness-of-fit test afterwards to get reasonable confidence.

-Ben Leimkuhler