GNU Scientific Library Reference Manual - Third Edition (v1.12) by M. Galassi, J. Davies, J. Theiler, B. Gough, G. Jungman, P. Alken, M. Booth, F. Rossi Paperback (6"x9"), 592 pages, 60 figures ISBN 0954612078 RRP £24.95 ($39.95) Get a printed copy>>>

GNU Scientific Library Reference Manual - Third Edition (v1.12) Buy the book here >>> support free documentation

36.1 Overview

Least-squares fits are found by minimizing \chi^2 (chi-squared), the weighted sum of squared residuals over n experimental datapoints (x_i, y_i) for the model Y(c,x),

\chi^2 = \sum_i w_i (y_i - Y(c, x_i))^2

The p parameters of the model are c = {c_0, c_1, ...}. The weight factors w_i are given by w_i = 1/\sigma_i^2, where \sigma_i is the experimental error on the data-point y_i. The errors are assumed to be gaussian and uncorrelated. For unweighted data the chi-squared sum is computed without any weight factors.

The fitting routines return the best-fit parameters c and their p \times p covariance matrix. The covariance matrix measures the statistical errors on the best-fit parameters resulting from the errors on the data, \sigma_i, and is defined as C_{ab} = <\delta c_a \delta c_b> where < > denotes an average over the gaussian error distributions of the underlying datapoints.

The covariance matrix is calculated by error propagation from the data errors \sigma_i. The change in a fitted parameter \delta c_a caused by a small change in the data \delta y_i is given by

\delta c_a = \sum_i (dc_a/dy_i) \delta y_i

allowing the covariance matrix to be written in terms of the errors on the data,

C_{ab} = \sum_{i,j} (dc_a/dy_i) (dc_b/dy_j) <\delta y_i \delta y_j>

For uncorrelated data the fluctuations of the underlying datapoints satisfy <\delta y_i \delta y_j> = \sigma_i^2 \delta_{ij}, giving a corresponding parameter covariance matrix of

C_{ab} = \sum_i (1/w_i) (dc_a/dy_i) (dc_b/dy_i) 

When computing the covariance matrix for unweighted data, i.e. data with unknown errors, the weight factors w_i in this sum are replaced by the single estimate w = 1/\sigma^2, where \sigma^2 is the computed variance of the residuals about the best-fit model, \sigma^2 = \sum (y_i - Y(c,x_i))^2 / (n-p). This is referred to as the variance-covariance matrix.

The standard deviations of the best-fit parameters are given by the square root of the corresponding diagonal elements of the covariance matrix, \sigma_{c_a} = \sqrt{C_{aa}}. The correlation coefficient of the fit parameters c_a and c_b is given by \rho_{ab} = C_{ab} / \sqrt{C_{aa} C_{bb}}.