From: Hans D Mittelmann <mittelmann@asu.edu>
Newsgroups: sci.math.num-analysis
Subject: Re: Least squares fit to sinusoidal data
Date: Tue, 04 Nov 1997 19:30:44 -0700

Marcus wrote:
> 
> I have some collected data which I know theoretically should fit a
> sinusoid.  I need to write an algorithm to do this.  I am familiar with
> some linear regression techniques, but this one is non-linear, I think.
> Could you tell me how to go about deriving the necessary equations to fit
> the data?  And, can I use some kind of modified least squares method?  This
> seems like a rather common thing to do in data analysis, but I am having
> trouble finding the information necessary to derive the formulas.
> The parameters I require from the data are the terms a,b,c,d in the
> equation below.
> y(x)=a+b*sin(c*x+d)
> 
> Thanks,
> Marcus
Hi,
for your purposes orthogonal distance regression may be the best
approach and an excellent program for that is ODRPACK. A link to it and
to GUI for it is at (go down a bit from there)
     http://plato.la.asu.edu/guide.html#lsq
If you prefer C, the next link is to a C code.

-- 
Hans D. Mittelmann			http://plato.la.asu.edu/
Arizona State University		Phone: (602) 965-6595
Department of Mathematics		Fax:   (602) 965-0461
Tempe, AZ 85287-1804			email: mittelmann@asu.edu