Generating Linear and Quadratic Cox Mixture Models

Abstract

While Scheffé canonical models have much to offer in the analysis of mixture data, their strong point does not lie in the ease with which effects of mixture components can be inferred from parameter estimates. The motivation for Cox mixture models, which are overparameterized versions of Scheffé models, is that inferences are easily made from Cox parameters about gradients and curvature along Cox-effect directions. Given a Scheffé model, there are an infinite number of equivalent Cox models, but once a reference blend, or base point, is specified, there is a unique Cox model. A FORTRAN program is described that will calculate linear and quadratic Cox polynomials for 2 to 10 mixture components. Required inputs are the design matrix, responses, and the composition of base points of interest.

• Blending
• Mixture Models
• Polynomial Models
• Response Surface Methodology

Additional information

Notes on contributors

Wendell F. Smith

Dr. Smith recently retired from his position as a Senior Research Associate in Imaging Research and Advanced Development.

Todd A. Beverly

Mr. Beverly is a System Administrator in Imaging Research and Advanced Development.