Fig. 5.

The effect of w on the empirical bias for a pair of related sequences as a function of the window size. Here, $r_1=0.1$, L = 10 000, k = 16, $w\in \{20,100,200,\dots ,1000\}$, and there are 50 mutation replicates.