Automated Fitting of Nonstandard Models

Abstract

A method for automated parameter estimation and testing of fit of nonstandard models for mean vectors and covariance matrices is described. Nonlinear equality and inequality constraints on the parameters of the model are allowed for. All the user will need to provide are subroutines to evaluate the mean vector and covariance matrix according to the model and, if required, the constraint functions. Subroutines for derivatives need not be provided. Some applications are described.