SOME OPTIMUM REGRESSION DESIGNS FOR DISCRIMINATING BETWEEN TWO MODELS.
Abstract
Very often two or more mathematical models are proposed for relating the system performance to selected variables which may affect the performance. In order to obtain the correct model statistical techniques must be provided for discriminating between models and for designing an experiment which minimizes the probability of selecting the incorrect model. Because of convenience in analyses and partly because of precedence experimenters frequently select designs which require an equal number of observations at each of several equally spaced levels. However, the results of this paper indicate that for purposes of discrimination between models the best allocation of the experimental points and the numbers of observations is not the one typically used. Many of the results given herein apply to the discrimination between a hyperbolic and a linear model; however, the techniques and some of the results have general application to other model forms. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 27, 1967
- Accession Number
- AD0649931
Entities
People
- A. C. Nelson Jr.
Organizations
- RTI International