Table 1.
Rare variant association tests under comparison.
Fig 1.
Simulation of a population stratification in case-control data.
A. Simulated demographical model with the cosi program. Modifications from the bestfit model designed by [34] are indicated in red. The migration parameter we varied is in dark blue. B. Geographical origin of cases and controls. In a first scenario cases and controls are from the same population A. In scenarios with a population stratification, the percentage of controls from population B we varied is indicated in dark blue.
Fig 2.
Comparison of FST values between simulations and real population genetic studies.
FST values obtained from simulations are plotted in function of the migration rate parameter. Pairwise FST values from two real population genetic studies, [40] and [41], are added respectively in red and blue.
Fig 3.
Type I errors at level α = 5% with a population stratification.
Bars represent type I errors without correction for population stratification. The red line corresponds to α = 5% and blue lines correspond to 95% confidence interval. Confidence interval is computed assuming that the number of false positives follows a binomial distribution with parameters 10,000 and 0.05. Correction methods are performed with the first two PCs.
Fig 4.
Powers at α = 5% after PCA model correction.
PCA model correction was performed on all scenarios except when 100% of controls are from population B (geographical covariates may totally predict the phenotype).