In addition to power (and type I errors), there are other characteristics that are important to take into account when deciding upon the appropriate multivariate GWAS analysis. MV-PLINK output results contain trait loadings, which indicate how much each trait contributed to the multivariate association result [7]. MV-BIMBAM outputs marginal posterior probabilities for each trait being unaffected, directly affected or indirectly affected by the QTL, conditional on an overall association with at least one trait [22]. PCHAT gives the weights for each of the traits included in the analysis which were used to construct the optimal linear combination of the traits to detect an association with the QTL [4]. MultiPhen output contains the betas and p-values for the association of each trait with the QTL based on the joint model including all traits [12]. This additional information, which is not provided by MV-SNPTEST and TATES, can be used to obtain insight into underlying biology and facilitates the discrimination between independent and pleiotropic QTL effects. Furthermore, MV-PLINK, MultiPhen, TATES and UV-MA allow analysis of a combination of quantitative and binary (case-control) traits
