| Abstract: | Correlation between adjacent plots in field experiments are common. In this paper the effect of correlation on the usual analysis of variance in a randomized block design is studied. A generalised analysis of variance where correlation is taken into consideration, and the analysis of covariance where the residuals of the neighbouring plots are used as covariates, is discussed. Uniformity trials are used as a basis of a Monte-Carlo study. The generalised analysis improved the power of the tests. The analysis of covariance method was not better than the usual analysis of variance method. |
