*4.4. Genomic Prediction Models*

Three statistical methods RR-BLUP [9,17,20], Bayesian LASSO (BL) [20,25,33] and Bayesian ridge regression (BRR) [25,72] were used to build GP models for PS. These predictive models estimate marker effects by modelling markers as random effects. No fixed effects were fitted in the models. The statistical models and their computation procedures are described in detail elsewhere [40,73]. The R package rrBLUP [56] was used to fit the RR-BLUP model and the R package BLR [74] was used to fit the BL and BRR models. The parameters used to fit BL and BRR were determined based on suggestions of de los Campos et al. [74]. Broad-sense heritability (0.25) of PS estimated in the population [3] was used. When preparing QTL marker data for model construction, the positive-effect allele of the tag SNP of a QTL was coded '1' and the alternative allele '−1'. Similarly for the SNP marker set, the reference allele of an SNP was coded '1' and the alternative allele '−1'. Missing data were coded '0'. The EM algorithm implemented in the R package rrBLUP [56] was used to impute the missing marker data because missing marker data were not allowed in the model construction.
