Figure 6

Scatterplot of differences in quartet support scores and topological error of wQMC and ASTRAL-2 trees. Each point (x, y) represents a dataset in which wQMC produced a tree with quartet support score x points higher than produced by ASTRAL-2, and with tree topological error y points lower. All values of x are strictly positive (we are only showing cases where wQMC produces a better quartet support score than ASTRAL-2), but values of y can be arbitrary. Points with y < 0 indicate datasets where ASTRAL-2 produces a topologically more accurate tree than wQMC, points with y = 0 indicate datasets where ASTRAL-2 and wQMC produce trees of equal accuracy, and points with y >0 indicate datasets where ASTRAL-2 produces a tree that is topologically less accurate than wQMC. Of the points that are not on the y = 0 line, more are below the y = 0 line than above (i.e., 12 below compared to 9 above), indicating that ASTRAL-2 tends to produce more accurate tree topologies than wQMC on these datasets. Also, when wQMC is more accurate, the improvement is lower than when ASTRAL-2 is more accurate. Thus, even when wQMC finds trees with better quartet scores, ASTRAL-2 tends to produce more topologically accurate trees. Plots in the margins are histograms of the x− and y−axes.