Fig. 9

SVDquartets vs. NJMerge given SVDquartet constraint trees and average gene tree internode distance (AGID) matrix. Subplots on top row show species tree estimation error (defined as the normalized RF distance between true and estimated species trees); note that gray bars represent medians, gray squares represent means, gray circles represent outliers, box plots are defined by quartiles (extending from the first to the third quartiles), and whiskers extend to plus/minus 1.5 times the interquartile distance (unless greater/less than the maximum/minimum value). Subplots on bottom row show running time (in minutes); bars represent means and error bars represent standard deviations across replicate datasets. NJMerge running times are for computing the subset trees “in serial”; see Eq. (1) in the main text for more information. The numbers of replicates on which the methods completed is shown on the x-axis, e.g., \(N=X,Y\) indicates that SVDquartets completed on X out of 20 replicates and that NJMerge(\({\mathcal {T}}_{SVD},D_{AGID}\)) completed on Y out of 20 replicates. SVDquartets did not run any datasets with 1000 taxa due to segmentation faults