Assessing Stability of Crop Varieties with Incomplete Data

Abstract

Joint regression has been very popular among plant breeders to evaluate stability of crop varieties tested under multi location-year trials. The plant breeders often finish their investigation for stability with Eberhart and Russell model (1966) though the component of deviation from linear regression is found significant for many varieties. Consequently,in such situations one cannot do ranking of all the genotypes tested with regard to stability. Eventually pair—wise comparisons with respect to stability can be made only in a subset of genotypes, whose deviations from linear regression are found not significant. This paper aims to emphasize the limitations of Eberhart and Russell model in evaluating stability of a set of varieties tested and suggest plant breeders alternative measures of stability when Eberhart and Russell model fails to comment on the stability of a sufficiently large number of varieties tested. Another problem of plant breeders that this paper also attempts, is dealing with the situation of the stability analysis when some cells in two-way table of genotype >< environments are blank. This paper examines methods cited in literature to handle incomplete data situations and brings out their practical relevance in the current generation of computers.An attempt has been made to develop handy computationalalgorithms wherever required and compares various procedures with respect to their capabilities in evaluating stability of the varieties