Extended Data Fig. 3: Comparison of proportional Robinson-Foulds distances of inferred trees from the correct simulated trees.

On the Y axis we show the proportional Robinson-Foulds distances (that is, normalized by 2(m − 3) with m the number of samples in the tree) of the tree estimated by each method with respect with the true simulated tree of the corresponding scenario and replicate. We collapsed tree branches of the simulated trees where no mutation event was simulated. Trees were compared as unrooted, and polytomies were compared as such (we collapsed branches of inferred trees with length equal to the minimum allowed length by the corresponding inference method). A Results for simulated datasets with no rate variation or ambiguity. B Results on simulated data with rate variation but no ambiguities. C Simulated data with no rate variation but with ambiguities. Other details are the same as in Extended Data Figure S1.