Figure 5.—

Comparison of the strict branch-site test, the 50:50 mixture test, and the empirical test on simulated sequences. A total of 215 sequences were simulated under a model of positive selection using a d_N/d_S value for the foreground lineage of 2 (yellow, orange, and red bars) or 10 (green, blue, and black bars). The percentage correctly classified was calculated using three different methods of assessing significance: (1) the strict branch-site test, which uses a -distribution, (2) a 50:50 mixture of a point mass of 0 and the , and (3) the empirical test. For the empirical test, the threshold for each branch length was defined on the basis of the distribution of LRS values for the paired neutral simulation (foreground d_N/d_S = 1).