Extended Data Fig. 8: Comparing genetic drift shared with PNG using F-statistics.

A. p-values from a Mann Whitney U test performed pairwise between the Australian samples grouped by community. Here we assume that the outgroup F₃ statistic for each individual (relative to PNG) in Fig. 3a is drawn from a common distribution for each community. The distributions of the statistic for a pair of communities are compared using the Mann Whitney test. Significant p-values (less than 0.05; shown in red) indicate the null hypothesis, that the distributions of the statistic for each group are equal, has been rejected. A two-sided test was used, with the Bonferroni p-value adjustment method. B. Matrix of all pairwise \({{\rm{F}}}_{4}^{({\rm{T}})}\) statistics (calculated using ADMIXTOOLS) of the form \({{\rm{F}}}_{4}^{({\rm{T}})}\) (YRI, PNG; X, Y), where ‘X’ and ‘Y’ are any one of the Australian populations in the NCIG dataset. Here we separate the Tiwi samples into the islands they are sampled from. Numbers reported are Z scores (the default ADMIXTOOLS output) and are significant when they exceed +3 or −3. See Methods for description of ‘Tiwi_Outlier’ label. C. Table showing all possible F₃ statistics of the form F₃(AUAx; PNG, AUAy), where ‘AUAx’ and ‘AUAy’ (simply labelled ‘X’ and ‘Y’ in this figure) are a pair of groups from the NCIG dataset. Here we separate the Tiwi samples into the two islands, Bathurst and Melville, and we also treat the Tiwi Outlier individuals as a separate group (See Methods for a description of the Tiwi Outlier individuals and a justification for removing them from our main analyses due to evidence of substantial recent admixture with PNG in their genomes). Text within each cell is the Z-score for the F₃ statistic from a block jackknife (directly from the software Admixtools). Following the theory of Patterson et al. (2012), statistically significant evidence of admixture between PNG and AUAx, but not AUAy, is indicated by a Z-score lower than −3, here indicated by red.