Note on the Missing Plot Procedure in a Randomized Block Design

Abstract

The properties of the missing plot procedure for a randomized block design are investigated from the randomization viewpoint. Expected mean squares, estimates of treatment effects, variances, and estimates of treatment effect variances are obtained when a particular block-plot combination is missing from the conceptual population of yields, when a block-treatment combination is missing, and when a restricted set of randomizations is considered. With proper definition of parameters and with certain assumptions, the analysis of variance is shown in all cases to lead to an unbiased test. Comparison with the infinite model approach is stressed.