GENOTYPIC EVALUATION WITH CROSSING WHEN THERE IS NESTING BETWEEN YEARS AND LOCATIONS

Abstract

The importance of the relationship between model and experimental situation which is to be described has been recognized since the origin of the principles and basic theory of the general linear model. In the genotypic evaluation it is advisable to make an experiment in each combination of a years (A) and I locations (L). A model that could be used for this series of experiments is the one based on the consideration that A is nested by L [model (3)]. Another model, frequently used, is based on the consideration that A is crossed with L [model (2)]. In this study, the effect of using model (2) when A is nested by L is studied. The study was mainly devoted to the estimation of the variance components for breeding purposes. With respect to the genetic variance its estimator was unbiased when the variance of the AxG interaction was the same in each location. The estimator of the variance of the GxL interaction was always unbiased, and the estimator of the general interaction AxG showed an expected value equal to zero. Model (2) estimators of the effects G and GxL, were; as those from model (3), unbiased, but their variances were larger. These larger variances must increase the probability of making wrong decisions in the planning or execution of plant breeding programs.