SEARCH

SEARCH BY CITATION

FilenameFormatSizeDescription
evo12193-sup-0001-supmat.pdf8362K

Figure S1. Distribution of Pearson's correlations between each posterior predictive simulated data set and the observed data, highlighting the improved fit of the overdispersion model to describe: (A) the Teosinte data set and (B) the HGDP data set.

Figure S2. Map of teosinte populations sampled, colored by their median estimated population-specific overdispersion parameter, inline image.

Figure S3. Map of human populations included in the analysis, colored by their median estimated population-specific overdispersion parameter, inline image.

Figure S4. Histograms of P-values produced by the partial Mantel test (with 1,000,000 permutations) on the 140 data sets for which the true contribution of ecological distance to genetic differentiation was 0.

Figure S5. Trace plots of the α parameters of the covariance matrix Ω.

Figure S6. Joint marginal plots of the α parameters of the covariance matrix Ω, colored by the MCMC generation in which they were sampled.

Figure S7. Acceptance rates for the parameters of the model that are updated with random-walk samplers, plotted over the duration of an individual MCMC run. Dashed green lines indicate the bounds of acceptance rates that indicate optimal mixing: 20%-70%.

Figure S8. Heatmapped matrices showing the performance of the model at all pairwise population comparisons. The posterior predictive p-value was defined as 1-2× | 0.5-ecdf(FSTobs)|, in which ecdf(FSTobs) is the empirical cumulative probability of the observed FST between two populations from a distribution defined by the posterior predictive sample for that population comparison, representing the p-value of a two-tailed t-test. Higher p-values indicate better model fit. Populations are enumerated on the margins, and may be referenced in SuppMat Table 1. a) The standard model. b) The overdispersion model.

Figure S9. Heatmapped matrices indicating the performance of the model at all pairwise population comparisons. The posterior predictive p-value was defined as 1–2 ×|0.5-ecdf(FSTobs)|, in which ecdf(FSTobs) is the empirical cumulative probability of the observed FST between two populations from a distribution defined by the posterior predictive sample for that population comparison, representing the p-value of a two-tailed t-test. Higher p-values indicate better model fit. Populations are enumerated on the margins, and may be referenced in SuppMat Table 2. a) The standard model. b) The overdispersion model.

Figure S10. Trace plots of the marginal posterior estimates for the αED ratio from MCMC analysis of the teosinte dataset. Inset figures give the marginal densities and 95% credible set for the samples after a burn-in of 20% a) The standard model. b) The overdispersion model.

Figure S11. Trace plots of the marginal posterior estimates for the αED ratio from MCMC analysis of the HGDP dataset. Inset figures give the marginal densities and 95% credible set for the samples after a burn-in of 20% a) The standard model. b) The overdispersion model.

Table S1. Metadata for populations used in the teosinte data set.

Table S2. Metadata for populations used from the HGDP data set.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.