Re: 3D Regression

Article View: sci.math.num-analysis

Article #275

Back to sci.math.num-analysis Previous Next

#275

From: W.A.B.Evans
Date: Sat, 11 Apr 1992 11:39

22 lines
1461 bytes

In article <44424@mentor.cc.purdue.edu> vbl@mace.cc.purdue.edu (Randall Julian) writes:
>I am interested in fitting a 3D data set to a user specified equation.  The
>equation is linear (Legrendre Polynomials) in the coefficients.  _Numerical
>Recipes in C_ has no routines of this type, quitting at 2D, as most references
>do.  If anyone has a reference to some C code for this please post, or
>e-mail me.
>
>|   Randy Julian                              |   Purdue University           |


	Surely, what you need here is EXACTLY the usual LINEAR regression
code but with the (trivial) modification that the fitfunction be defined as
a function of NV variables (2 or 3 in your case) rather than 1. All other
aspects are (I believe) unaltered.

	My own codes for (Linear and Nonlinear) Regressions (in FORTRAN)
have this option built in (the default being NV=1) - and have served me
perfectly well for the odd occasions when I needed a least squares fit to
a function of more than 1 variable.

						W. Alan B. Evans
						[ wabe@ukc.ac.uk ]

Technical Details

Message-ID: <725@hobby.ukc.ac.uk>

Path: arc1-ca.pugleaf.net!sparky!uunet!mcsun!uknet!harrier.ukc.ac.uk!hobby.ukc.ac.uk!wabe

References: <44424@mentor.cc.purdue.edu>

Show Full Headers

Complete Article Headers

Path: arc1-ca.pugleaf.net!sparky!uunet!mcsun!uknet!harrier.ukc.ac.uk!hobby.ukc.ac.uk!wabe
From: wabe@ukc.ac.uk (W.A.B.Evans)
Newsgroups: sci.math.num-analysis
Subject: Re: 3D Regression
Message-ID: <725@hobby.ukc.ac.uk>
Date: 11 Apr 92 11:39:19 GMT
References: <44424@mentor.cc.purdue.edu>
Reply-To: wabe@ukc.ac.uk (W.A.B.Evans)
Organization: Computing Lab, University of Kent at Canterbury, UK.
Lines: 22

Reply to Parent Message