The cumulative effect of genetic markers on classification performance: Insights from simple models

► We describe a mathematical generalization of a simple model for explaining the cumulative effect of markers on classification. ► An alternative model for explaining the effect of additional loci is based on an optimal classifier. ► Cases in which the expected performance does not profit from additional loci are characterized in simple terms. ► Further insight comes from an information-theoretic perspective. ► Multilocus dependency measures are shown to tightly correspond with misclassification rates.