Null simulations.

Observed type I errors at selected significance levels for the Generalised C-alpha test for association with a quantitative trait and a dichotomised version of a quantitative trait in a 50 kb region where the rare variants tested do not account for any of the trait variance. Tests only consider variants in the region with a maximum MAF of 1% and a minimum MAF as indicated in the table. Type I error is estimated over 100,000 replicates of data for a sample of size 2,000. Significance in each replicate of data is assessed empirically by random permutation of the quantitative trait value and recalculation of the test statistic 1,000 times as described in Text S1.