Figure 3.

Comparison of the −log₁₀(p) test statistics of MIVQUE(0) (minimum variance quadratic unbiased estimation) with REGULAR (regularized F test) from the real data analysis (A) and from the analysis of a permuted sample (B). The Pearson correlation coefficients between the test statistics of the two methods are represented by $r_{xy}.$ These plots demonstrate that the test statistic of the mixed model is highly correlated with the test statistic of the regularized F test, but the patterns are different for the real data and the permuted data.