A General Lease-Squares Fitting Package for the Sun Workstation

This document introduces a user interface that provides fast and efficient general least-square analysis of experimental data. It is suitable for use in the Sunwindows environment on Sun UNIX workstation. The package is capable of providing solutions of both linear and non-linear least-squares models of experimental data, and incorporates full graphics and error analysis. Known uncertainties in the data are easily incorporated as weights for each experimental point if required. An overview of the operational details of the package is given in Section 2. Examples of how to use the package are given in Section 3. The linear least-squares analysis allows unweighted, y-weighted and x - and y-weighted optimisation, and in the former cases either the slope or intercept may be fixed if required. Since the theory of unweighted and y- weighted linear least-squares is well known 1, 2, it will not be described in this document. The theory for the case where both the dependent and independent variables exhibit uncertainties is not well known however, and as such is outlined in Appendix A. Appendix B details the theory of non-linear least- squares optimisation using the Levenberg-Marquardt algorithm, which is the routine implemented in this package.