Ten years ago, with @cfraisserios, we were dissatisfied with averaging windows in genome scans and decided to calculate FST90 (the 90th percentile of the FST distribution) and FSTmax (the maximum FST of the window) instead. We struggled to convince the reviewers that this was a good idea. We now have a paper we can cite to justify that it's the right approach. Why FST90 in addition to FSTmax? Because in true data FSTmax is full of rotten SNPs. #PopGen

https://doi.org/10.1093/gbe/evac143