CV-?: designing validations sets to increase the precision and enable multiple comparison tests in genomic prediction

Usually, the comparison among genomic prediction models is based on validation schemes as Repeated Random Subsampling (RRS) or K-fold cross-validation. Nevertheless, design of training and sets has a high effect way subjectiveness we compare models. Those procedures cited above have an overlap across replicates that might cause overestimated estimate lack residuals independence due to resampling issues less accurate results. Furthermore, ANOVA multiple-comparison tests, such Tukey, are not recommended assumptions unfulfilled regarding residuals' independence. Thus, propose new sample observations build cross-validation alpha-based (CV-?). The CV-? was meant create several scenarios (replicates x folds), regardless number genotypes. Using CV-?, genotypes in same fold much lower than cross-validation, indicating higher residual Therefore, results, proof concept, via ANOVA, could proposed methodology RRS applying four with simulated real dataset. Concerning predictive ability bias, all methods showed similar performance. However, mean squared error coefficient variation, method presented best performance under evaluated scenarios. Moreover, it no additional cost complexity, more reliable allows non-subjective factors. can be considered precise for model selection.