Analysis of variance of a randomised block design with missing observations.