Resumen de Bayesian Variable Selection for Fractional Factorial Experiments with MultilevelCategorical Factors

Phil Woodward, Rosalind Walley

• The Bayesian approach to fractional factorial experiments is useful for its ability to identify difficult-to-spot interaction effects in experiments with complex aliasing. However, experiments using this approach have consisted either of two-level factors or factors with quantitative levels with priors that cannot be used in experiments with multilevel categorical levels. A simple multivariate normal prior is presented that is a natural extension of the prior used for two-level factors. Its effectiveness is demonstrated by reanalyzing two multilevel factorial experiments with complex aliasing.