Ascertainment correction for a population tree via a pruning algorithm for likelihood computation

Abstract

We present a method for correcting ascertainment-bias in a coalescent-based likelihood for population trees. Our method is computationally simple and fast. To correct for the bias we compute the probability of allele-counts conditioned on the locus being included. This conditional probability is simply the uncorrected likelihood divided by the inclusion probability. A modification of a pruning algorithm is introduced so that the inclusion probability can be computed with a single run of the algorithm. Our computation is exact and avoids Monte-Carlo based methods.