Test for changing sex-bias applied to bottleneck simulations with a proportion of females of
p1 outside the bottleneck and
p2 during the bottleneck. (A) Off-diagonal entries are power: values greater than 0.85 are in blue, values between 0.85 and 0.10 are in light yellow, and values less than 0.10 are in orange. Entries on the diagonal are false positive rates and are in gray. (B, C) Data was simulated with a female bias outside the bottleneck (
p1 = 0.8) and a varying proportion of females during the bottleneck,
p2. Our method produces separate estimates for
p1 and
p2 (
, in blue), while the the
π-based estimator produces one estimate (
pπ, in red). Bias of estimators of the proportion of females (B) outside the bottleneck,
, and (C) during the bottleneck,
, is shown for both methods. The
π-based estimator is biased when
p1 and
p2 differ (here, when
p2 is not 0.8), whereas our estimator is unbiased.