library(discent)
Current implementation scales approximately \(O(n^2k)\) where \(n\) and \(k\) are number of demes and average samples per deme pair, respectively. but model is fast
bootstrap with traditional approach