Abstract: (This is on a joint work with Dr. Leonid Khinkis.)
Current methods for least squares regression analysis of
nonlinear models can often lead to unusual results,
including multiple minima. This diminishes the credibility
of the associated estimates of the model parameters and
their confidence limits. In this presentation, the
characteristics of nonlinear models and nonlinear regression
analysis will be compared to the corresponding
characteristics of linear models. Preliminary results from a
new method will be presented which address many of the
shortcomings of traditional nonlinear least squares
regression analysis.