Accuracy of admixture graph inference
Evaluating complex admixture graph inference from f-statistics through simulation
Admixture graphs inferred from f-statistics are widely used to reconstruct complex histories of population and species mixture. I evaluated the accuracy of these inference methods on more complex graphs than have typically been tested, using extensive simulations.
The results show that accuracy drops considerably as graph complexity increases, which has implications for how confidently these methods can be applied to real, complicated evolutionary histories. See (Frankel & Ané, 2026) for the full study.
