A unified theory of missing plots in experimental designs (Q788445)

This paper provides a unified account of missing plots in experimental designs. The authors give an account of the value to be substituted in the missing plot, the increase in the variance of a treatment comparison, the increase in the average variance, and bias for the situations when we are dealing with balanced incomplete block designs, group divisible designs, randomized block designs, partially balanced incomplete block designs, linked block design, and also for a general incomplete binary design. The results for one or more missing observations are obtained using the general linear model, and the results also cover the non-full rank case.