Figure 3

Results for incompatible input trees and n = 32. The distances are normalized by dividing by the maximal distances, that is, 2n - 6 for the RF distance and n - 3 for the other two distances. "Missing splits" is the proportion of splits that are in the model tree but not in the supertree and "Incorrect splits" are in the supertree but not in the model tree. The mean of the data is displayed by "+". Note the different scaling of the y-axes.