We propose graphical methods for displaying relevant information on a selected parameter from a normal nonlinear regression model. It is shown that the usual extension of added variable plots from linear to nonlinear regression can fail to reveal important diagnostic information, and that this information can be recovered by using a parameter plot that depends on selected elements of the parameter-effects curvature array.
Bibliographical noteFunding Information:
This work was supported by grants from the National Science Foundation and the Monsanto Company. Kinky Larntz, Chih-Ling Tsai and the referees provided useful comments on an earlier version of the manuscript. Computational assistance was provided by Roy St. Laurent.
- Added variable plot
- Graphical method
- Parameter-effects curvature