Robust Nonlinear Least Squares Approaches for Evaluating OVA-Mediated Bleaching Reactions: An Experimental Comparative Study

Abstract

Nonlinear models are usually encountered in various areas including experimental studies such as physics, chemistry, biology etc. Ordinary least squares is one of the most widely used methods for parameter estimation in different types of nonlinear models. However, there are some regression assumptions need to be satisfied for obtaining efficient parameter estimates. In this paper, the parameter estimation process is evaluated carefully for some bleaching reactions by using chicken egg albumin (OVA) and some precautions are taken in the presence of violations of the assumptions (heteroscedasticity, autocorrelation, the presence of outliers). In this way, robust logged nonlinear least squares approaches are examined and compared under different conditions of reactions. It can be concluded that logged and robust analyses are preferable together in nonlinear regression in order to obtain efficient parameter estimates and reliable results. However, the best weight function in robust nonlinear least squares can vary for each condition.