Ridge regression and extensions for genomewide selection in Maize
This paper reviews properties of ridge regression for genomewide (genomic) selection and establishes close relationships with other methods to model genetic correlation among relatives, including use of a kinship matrix and the simple matching coefficient as computed from marker data. A number of alternative models are then proposed exploiting ties between genetic correlation based on marker data and geostatistical concepts. A simple method for automatic marker selection is proposed. The methods are exemplified using a series of experiments with test-cross hybrids of maize (Zea mays L.) conducted in five environments. Results underline the need to appropriately model genotype–environment interaction and to employ an independent estimate of error. It is also shown that accounting for genetic effects not captured by markers may be important.