Permutation procedures are time consuming, and an attractive alternative is to estimate an effective number of independent tests directly from the genotype correlation matrix. A moment–based estimator was recently proposed by Patterson et al. [2006] based on the eigenvalues of the correlation matrix. This estimator has good properties in the context of detecting population structure, but we wished to see whether it is equally useful for correcting multiplicity. Let (5) where n is the number of markers and M is a normalized matrix of genotypes with one row per subject and one column per marker [for details, see Patterson et al., 2006]. Denoting the eigenvalues of X by λ1,…,λm, where m is the number of subjects, the effective number of tests is estimated by (6) As for the permutation test, we estimated nP for a grid of subsampling densities. We calculated nP for each chromosome and summed to obtain a genomewide estimate.
