ABSTRACT

In this chapter, the authors focus on an experiment involving four small data sets, each of which is to be fit to two nonlinear sigmoidal growth models—namely, a Gompertz model and a logistic model. In addition, the nonlinear models were fit with and without reweighting. The difficulties in fitting a data set to a hypothesized model are well known when the model is linear, especially when there is evidence of one or more outliers in the data. When the model to be fit is nonlinear in the parameters, the problem is further complicated by the numerical difficulties one can encounter in trying to obtain the best estimates for the model’s parameters. Because nonlinear regression problems are essentially nonlinear programming (NLP) problems, all of the known difficulties in solving NLP problems are present in these situations.