A new uniform mapping algorithm for sample selection

A new algorithm for sample selection in the field of experimental design is presented. This method was designed for a multivariate data set of samples where the variables' levels are pre‐determined by the system instead of being determined by the practitioner and especially appropriate for non‐symmetrical multivariable space. Symmetrical experimental designs are not applicable in either of the cases. Anyway, this method is also applicable to symmetrical designs. Several tests have been made for symmetrical and non‐symmetrical cases and also about an already treated real case, comparing the new algorithm with others previously known in literature. Its performance has been checked considering several evaluation quality criteria like D‐optimal, A‐optimal, G‐optimal and models error prediction. Copyright © 2008 John Wiley & Sons, Ltd. A new algorithm for sample selection in the field of experimental design is presented. This method was designed for a multivariate data set of samples where the variables' levels are pre‐determined by the system instead of being determined by the practitioner and especially appropriate for non‐symmetrical space. A dramatic difference in quality criteria was shown in favor of the new algorithm based on the furthest points to sequential centroids. As a result, the prediction errors for a general linear model are diminished.