Figure 3.

Mean difference of –log10p-values between the multivariate and univariate F-tests when phenotypes are missing. The p-values of the univariate F-test were obtained from the test on the first variate, which had the greatest QTL effect. Color scales are the same as those in Figure 1. d, r, and a_d denote the number of variates, phenotypic correlation, and the relative size of QTL effects between variates, respectively. m_prop denotes the missing proportion of phenotypes. At each d, the results when m_prop is the maximum (i.e., each individual has a phenotypic record only for a single variate) are presented in the last column. The results when d = 8 and m_prop = 0.656 (the third column on the bottom row) are not shown because of the high frequency of failure in variance component estimation. For results when d = 4 and d = 8, the mean differences at the scenarios that were not tested were interpolated using spline regression implemented in the mgcv R package.