File:Model validation using independent statistics for models of human origin evolution.webp

English: "a–c, Using our best-fit models, we simulated expected cSFS and compared the simulated spectra to those observed from the data. Our inferred models provide a good fit to the data, even though this summary was not used in our inference. Across the three populations (a, Nama; b, Mende; c, Gumuz), ancestral-state misidentification was consistently inferred to be 1.5−1.7% for intergenic loci (Supplementary Information section 6.2.2). d,e, We used Relate16 to reconstruct genome-wide genealogies, which we used to estimate coalescence-rate trajectories and cross-coalescence rates between pairs of populations. Although coalescence-rate distributions are informative about past evolutionary processes, interpretation can be hindered by migration and population structure, and translating RCCR curves into population divergence times is especially prone to misinterpretation. d, Real data; e, our model. In our model, the Mende–Gumuz split occurs before the Gumuz–British split. However, the model also predicts a recent elevated Mende–Gumuz RCCR. This pattern, also observed in the data, does not indicate that the Mende and Gumuz split more recently than the Gumuz and British populations."