Fig. 2
From: Weighted ASTRID: fast and accurate species trees from weighted internode distances
Topological error of species tree across methods on the ASTRAL-III S100 dataset (\(n = 101, \text {AD} = 46\%\)). Subfigures vary the sequence lengths, affecting the gene tree estimation error (measured in GTEE, the average distance between estimated gene trees and true gene trees). The x-axis varies in the number of genes. Results are shown averaged across 50 replicates with standard error bars. All weighted methods (wASTRID, wASTRAL) ran on gene trees reannotated with IQ-TREE aBayes support branch support and lengths. All methods achieve better accuracy when given more gene trees (larger k) or more accurate gene trees (lower GTEE). Weighted methods are more accurate than unweighted ones. wASTRID-s and wASTRAL-h have almost the same accuracy