A Selection Index to Cope with Genotype-Environment Interaction with an Application to Wheat Breeding

The objective of this study was to develop a genetic selection index for the combination of quantitative observations repeated in different environments when there may be genotype-environment interactions and heterogeneity of the variance. A distinction is made between the index for local ranking of strains for each environment on its own and the index for a joint ranking which predicts mean genotypic value averaged over all environments. The resulting index was applied to field trial observations on the yield of wheat at two locations, involving 276 newly developed strains in six trial groups of equal size, in order to study the efficiency of the index and the sampling variance of index weights. The index for local ranking of strains was found to be consistently more efficient than ranking without the index, though only marginally so in trials with low coefficients of error variation. Limitations on general use of the index, as related to trial size and the standard errors of index weights, are discussed.