Importance of the first design matrix in experimental simplex optimization
Tomas Öberg
Chemometrics and Intelligent Laboratory Systems 44, 147-151 (1998).

Abstract
The basic and modified simplex methods are efficient optimization techniques applied in many fields of chemistry and engineering. Various first design matrices were evaluated on polynomial models with added noise. D-optimal linear design matrices performed better than regular or cornered first simplices in the normal experimental situation. These findings have been implemented in a new experimental design an optimization software, the MultiSimplex®.

Postprint som en PDF-fil, 33 kb

DOI: 10.1016/S0169-7439(98)00161-0