Model Building and the Grid Technique
R. R. Scott, Edward A. Sylvestre · Journal of Quality Technology · 1979
Most of the literature on model building deals with either parameter estimation or model discrimination for prespecified models. The models under consideration are usually (1) developed by a mechanistic treatment of the problem, (2) available from previous treatments for the problem, or (3) assumed to be polynomials. Often one finds that the process under study is extremely difficult to characterize mechanistically, and an experiment is run to determine the form of the relationship. This paper will present a general technique that we have found to be very useful in helping to develop an empirical model. The technique attempts to develop the form of the multipredictor model by considering successive one-variable models.