Some practical tools for everyday least squares

A few issues often encountered in least-squares estimation are discussed. A simple and general test for the goodness of fit is proposed, and an adaptive criterion for testing the convergence of iterative nonlinear problems is suggested. The old problem of the number of significant digits of correlated parameters is addressed and a new algorithm for rounding the numerical values is described. The underlying principles are emphasized and the accent is on practical implementation of improvements to computer codes for data analysis. (C) 2001 Academic Press.