The problem of plot size in experiments with trees

Abstract

This paper develops a theory on the determination of optimum plot size, based on the intraclass correlation coefficient (ρ). This method has the advantage of using a well known statistic, which belongs always to the closed interval [- 1/(k - 1) < ρ < 1], where k is the number of units (subplots) within each plot. The plot taken as optimum is that with k units, which leads to minimum variance of treatment means, when the total area per treatment is constant. Guard rows are taken in consideration in the solution. The theory is specially appropriate for research on plot size in experiments with trees, with or without guard rows.