Identifying significant effects in fractional factorial experiments

Abstract

AbstractThis paper presents a method for identifying significant effects in fractional factorial designs. The method is based on comparing functions of the estimated effects with a scale invariant reference distribution. The method is illustrated and compared with alternative techniques on a dataset from processed cheese production. The method corresponded well to other methods (probability plotting and Bayes plot) that do not make use of error estimates. These methods were better than F‐tests with few (two or four) degrees of freedom for error.