gaussfit -- Least squares and robust estimation program.
gaussfit model environment
The gaussfit task solves least squares and robust estimation problems. It includes a programming language for formulating the estimation problems, a compiler and interpreter, and an algebraic manipulator for analytically calculating the required partial derivatives.
This task was developed by William H. Jefferys, Barbara E. McArthur, James McCartney, and Mike Fitzpatrick. It is an IRAF foregin task that is meant to be run from the host level and not from within the cl. The task and its programming language are described in detail in a separate manual that can be obtained from the STSDAS group at STScI.
The programming language provides an easy way to formulate general nonlinear problems; problems where the equation of conditions contain more than one observation; problems with correlated observations; problems where exact constraints among parameters must be enforced; and certain robust estimation methods that generalize least squares to non-Euclidean metrics. It also provides greater immunity against outliers than does the classic least squares method.