Table 1.
Detailed overview of the different scenarios considered for the type I error and power study.
Fig 1.
Quantile-quantile plots of SMT statistic values for singletons, doubletons, and all SNVs.
Datasets were generated from the null model described in scenario 0 in Table 1 of size n = 1,000 for m = 10,000,000 replicates. Quantile-Quantile plots are shown comparing the empirical quantiles of the t-test statistics of singletons (left panel), doubletons (middle panel), and all SNVs (right panel) to the theoretical quantiles of the tdf = 1000–4 distribution. For computational purposes, each plot is based on a random sample of 1,000,000 t-test statistics, out of the 132,797,000 t-test statistics of all singletons in all replicates, out of the 41,341,000 t-test statistics of all doubletons in all replicates, and out of the 325,393,000 t-test statistics of all SNVs in all replicates. In grey ribbons, approximate 95% point-wise confidence intervals are shown.
Table 2.
Empirical type I error of MMTs and SMT for different nominal α levels.
Fig 2.
Power estimates of the SMT and MMTs.
Data was generated under an alternative-hypothesis model described in scenarios 1–36 in Table 1 of size n = 1,000 for m = 10,000 replicates. The nominal α was set to 0.05 (upper panel) and 2.5∙10−6 (lower panel). In the lower panel with α = 2.5∙10−6, the coordinate system is shown on a log10-scale to better visualize the small power differences between the approaches. Multiple testing corrections for the SMT of all SNVs in a gene were done using the BH-correction.
Table 3.
Gene-level p-values for the association tests of candidate genes with SBP in the genetic analysis 19 data analysis.