Sequential analysis of variance for irregular designs: some Monte Carlo results

Abstract

Sequential analysis of variance was first dealt with by Johnson (1953) for equal numbers in groups and the results were extended to unequal numbers by Jones (1968) and Jones and Hinkelmann (1969). The average sample number (A.S.N.) obtained for an irregular design is usually lower than that for equal subclass numbers and this advantage is enhanced if sampling starts with the design. In this article Monte Carlo results are given for two classifications models with fixed effects, namely, one-way classification and twoway classification without interaction. Different methods are adopted to obtain the minimum starting design for unbalanced models and the resulting A.S.N. of these are compared. In all cases the probabilities of errors of the two kinds are taken to be 0.05.