Compensating for site and year effects in estimating the true yield of wheat genotypes

Abstract

When data are obtained for several cultivars over years and sites, it is possible to calculate three regression coefficients for each cultivar, based on site x cultivar effects, year x cultivar effects, and site x year x cultivar effects respectively. For wheat yield data from western New South Wales there is an approximate equality of these measures of phenotypic stability, and it is inferred that a single environmental factor has a dominant effect both between and within seasons. This result is applied to the estimation of a regression mean for each cultivar, a technique designed to compensate for the biased sampling of environments over one or two seasons. For nine wheat cultivars, it is shown that the regression mean is at least as good an estimator of true yield as the arithmetic mean, and the conditions under which a regression mean is expected to be a better estimator are defined.