EVALUATION OF GENOTYPES UNDER INTRAREGIONAL METEOROLOGICAL HETEROGENEITY. CONFOUNDING vs. NESTING IN YEARS AND LOCATIONS

Abstract

Frequently the evaluation of genotypes (G) in several locations (L) during several years (Y) is made under each level of a factor B in a randomized complete block design. In this study the analyses of variance (ANOVA) for the evaluation of rainfed annual crops are shown according to two cases: Y nested in L and Y deliberately confounded with L. Based on these analyses, the main objective was to determine the effect that the use of a model where L and Y are confounded when actually Y is nested in L, has upon the level of statisticcal significance of the F tests and on the estimators of the variances of G, B, and the interaction GB. These estimators showed positive biases involving the variance of the interactions GL, BL, and GBL, respectively. For the tests of the hypotheses for G, B, and GB, the ANOVA F values were greater and the values required to reach statistical significance decreased causing an increase in the assumed level of significance. In addition, only in the tests for Y/L (Y nested in L), B(Y/L), G(Y/L), BG(Y/L), and R/Y/L (R is “replicates”), differences were found between the use and the absence of split plots with B being the whole plot factor for random or mixed (when only G or B is a fixed-effect factor) model.