GaussFit—A system for least squares and robust estimation

GaussFit is a new computer program for solving least squares and robust estimation problems. It has a number of unique features, including a complete programming language designed especially to formulate estimation problems, a built-in compiler and interpreter to support the programming language, and a built-in algebraic manipulator for calculating the required partial derivatives analytically. These features make GaussFit very easy to use, so that even complex problems can be set up and solved with minimal effort. GaussFit can correctly handle many cases of practical interest: nonlinear models, exact constraints, correlated observations, and models where the equations of condition contain more than one observed quantity. An experimental robust estimation capability is built into GaussFit so that data sets contaminated by outliers can be handled simply and efficiently.Invited Talk at the 18th meeting of the Division on Dynamical Astronomy of the American Astronomical Society held in Braintree, Massachussetts, May 7–9, 1987.