Suppose m phenotypes are measured as indicators of one trait, e.g., individual symptoms within a disorder, items within a test, or multiple measures of one trait using different instruments (e.g., open-field test, a light-dark box, and an elevated plus maze to measure anxiety in mice). Rather than combining these m phenotypes into one general phenotype, we test the association between all m phenotypes and all n genotyped genetic variants (GVs) using a statistically appropriate method (e.g., linear or logistic regression). Let p(1)…p(m) be the ascending p-values of the m phenotypes for a given GV. TATES combines within each GV the m phenotype-specific p-values to obtain one overall trait-based p-value PT as follows:(1)where m e denotes the effective number of independent p-values of all m phenotypes for a given GV, and m ej the effective number of p-values among the top j p-values, where j runs from 1 to m, and p j denotes the j th p-value in the list of ordered p-values. PT is thus the smallest weighted p-value, associated with the null hypothesis that none of the phenotypes
