Improved Calibration of Nonlinear Mixed-Effects Models Demonstrated on a Height Growth Function

Abstract The calibration of a nonlinear mixed-effects model is of critical importance in making local predictions. In previous applications, simplified equations were generally used to calibrate nonlinear mixed-effects models fitted using the first-order methods implemented through the NLINMIX macro in SAS. This simplification, however, could distort local predictions. In this study, using a nonlinear height growth model of lodgepole pine (Pinus contorta var. latifolia Engelm.), we demonstrated on two data sets the procedures to obtain an improved calibration of the height growth model. The differences between the improved and conventional calibrations were found to be significant. Calibration of the nonlinear mixed models using the improved method resulted in not only reduced bias but also reduced variance of the errors. It is recommended that the improved calibration method be used. A computing program detailing the procedures for obtaining the improved calibration was developed.