Abstract

A wide variety of models have been proposed recently for experiments that vary the order in which components are added, or steps are performed. There are models with a linear effect of each component's position in the sequence, and models with added quadratic effects and product terms. Kriging models based on a component's position have also been proposed. Before models based on each component's position were introduced, models based on the relative position of each pair of components became popular. This talk connects these models and provides lack-of-fit tests in order to assist with model selection and interpretability of parameters. Examples illustrate the need for these varied models.