polyfitweighted

Simple function to calculate a polynomial best fit of 1D data Y with array of weights W (which must

Description

Like polyfit.m but includes weighting of each data point.

Summary
----------------
Find a least-squares fit of 1D data y(x) with an nth order polynomial, weighted by w(x).

Usage
-----

P = polyfitweighted(X,Y,N,W) finds the coefficients of a polynomial P(X) of degree N that fits the data Y best in a least-squares sense. P is a row vector of length N+1 containing the polynomial coefficients in descending powers, P(1)*X^N + P(2)*X^(N-1) +...+ P(N)*X + P(N+1). W is a vector of weights. Vectors X,Y,W must be the same length.

Class support for inputs X,Y,W:
    float: double, single

By SS Rogers http://ssr.phy.umist.ac.uk

• MATLAB release: MATLAB 7.3 (R2006b)

[John D'Errico]

Not bad. Uses a reasonable algorithm for estimation, although there is no real reason to use qr here if you do not intend to return parameter standard errors. Backslash should be appropriate here.

A minor point - why test for events that cannot possibly happen? For example why have a test for nargout>2, if there is only ONE output argument? The answer is the author has just copied the polyfit code including comments, and then modified it. I'd suggest that attribution would be nice when you copy code. At least clean it up.

I've chosen not to give any numerical rating to avoid any appearance of a conflict of interest (polyfitn also offers polynomial modeling.) But were I to offer a rating, it would not be a bad one.

[Antoni J. Canós Marín]

Have you seen the submitted function polyfit3.m (File Id. 4262)?

[Salman Rogers]

Thank you to John D'Errico for his comment - yes, this script is based on polyfit.m by The Mathworks which implements a well known algorithm. I have only added the weighting factor, and taken out the return of standard errors. polyfit.m is indeed acknowledged inside the file and in "help polyfitweighted". Apologies for any ambiguity.

[Matt J]

In the help comments for polyfitweighted2, it says that Z and W should be of size [length(X) length(Y)]

However, throughout the code (both polyfitweighted2 and polyval2, Z and W are treated as though of size
 [length(Y) length(X)].

[S]

Exactly what i was looking for