Comparing cultivar means in multilocation trials when the covariance structure is not circular

Abstract

Cultivar trials conducted at different locations are frequently evaluated by analysis of variance (ANOVA) techniques. One of the assumptions underlying such analyses is that residuals be independently distributed with common variance. Often, however, this assumption is violated. This contribution explores ways to analyse data when the usual assumptions regarding the variance-covariance structure do not hold. An example is given to demonstrate the proposed procedure. ©1996 The Genetical Society of Great Britain.