DESIGN OF A ROBUST ESTIMATOR FOR NONLINEAR KINETIC MODELING

A biased estimator for nonlinear kinetic modelling is studied. The pro posed estimator contains two easily tuned parameters that enable the e stimator to possess robustness in dealing with the problem of a nearly singular design matrix. The robustness of the estimator is discussed in detail, and the asymptotic normality of the estimator is proved con cisely but rigorously. The regularity conditions imposed to obtain the results are fairly weak in engineering applications. Three examples p resented demonstrate that the estimator is significantly better than t he least-squares (LS) estimator and the ridge estimator in both estima te accuracy and convergence speed.