Estimating Tajima’s D using ML estimates of the SFS. These two plots are based on neutral sets of scenario, each plotted data point is an estimate of Tajima’s D for a 1 Mb region. Subfigure ‘a)’ is high depth (8×), with an error rate of 1.0% and using a p-value of 10-6 and subfigure ‘b)’ is low depth (2×) with an error rate of 0.5%. For our full ML method in the high depth scenario all values fall in the vicinity of the y = x line, but shows higher variance for the low depth as expected. Notice that for the stringent cutoff both genotype calling methods overestimates.