Re: for help (Forwarded from Chuck Lindsey)

Brian E. Smith (smithb@MANAGEMENT.MCGILL.CA)
Tue, 26 Nov 1996 10:28:24 -0500

From: "Lindsey, Dr. Charles" <clindsey@fgcu.edu>
To: ' MATHEDCC' <mathedcc@vm1.mcgill.ca>
Subject: RE: for help
Date: Mon, 25 Nov 96 12:21:00 PST
Message-Id: <3299FB5E@smtpgate.fgcu.edu>
Encoding: 45 TEXT
X-Mailer: Microsoft Mail V3.0

Sorry to reply to the list, but there isn't a direct email address below.

I have had some luck with similar problems in the past just setting up a
spreadsheet to directly calculate the residuals and the
square error. You can experiment with value of the parameters and narrow the
optimal values down quite a bit, and then write a computer program to find
parameter values from there. It's crude but effective.

Chuck

Chuck Lindsey clindsey@fgcu.edu
University of South Florida at Fort Myers/
Florida Gulf Coast University
http://www.fgcu.edu/~clindsey/
------------------------------------------------------------------------

Hi Everyone,

In my research work, I got some experiment data, and
from the theorical ayalysis I believe that these data
fit an emprical formula which has a form as:
f(x,y)=cos(arctg(a*x**b+c*(x*y)**d))*exp(-e*(x**g)*(y**h))
where a,b,c,d,e,g,h are the parameters that should be determined,
and x>=0, y>=0, 0<=f(x,y)<=1.

I tried to do the regression analysis using some softwares,
such as GRAFTOOL, Matlab, Math CAD, while the form of the
above regression formula is too complicated to be analyzed
using these softwares.

I am writing to inqury if anyone have a computer program
suitable for my problem, and any other relative information
is also appreciated.

I am looking forward to receiving your early reply.

Sincerely yours

Yang Qingshan