All MV methods resulted in higher power than UV analysis. Power of MV-PLINK, MV-SNPTEST, MultiPhen and MV-BIMBAM was similar: between 10% and 15% for scenarios with a residual correlation of 0 or 0.3 and 20-25% for scenarios with rE = 0.7. PCHAT outperformed MV-PLINK, MV-SNPTEST, MultiPhen and MV-BIMBAM for scenarios with rE = 0 and 0.3, but not for scenarios with rE = 0.7, although power of PCHAT increased for rE = 0.7 as well. TATES showed a power between 11-14% for all scenarios and performed slightly better than UV-PCA and UV-MA, which showed a similar performance as UV analysis of trait 1, the trait associated with the QTL.
