Performance of the PIQMEE method on very large data sets. We analyzed data sets consisting of 12,000, 15,000, 18,000, 21,000, 24,000, 27,000, and 30,000 sequences with PIQMEE. All sequences were sampled at one point in time (homochronous sampling). (A) The chain length reached in 14 days. (B) The burn-in, expressed as the fraction of the entire chain, that needed to be removed in order to obtain well-mixing part of the chain. We note that if we have to remove more than half of the chain, it indicates that convergence is not necessarily reached and the chain should thus be run for longer. (C) The ESS reached. The y-axis in (C) has a log-scale.