Goodness-of-fit tests of five models of population history.

<p>Box plots indicate the posterior predictive distributions of the mean (a, c, e) and coefficient of variation (b, d, f) for nucleotide diversity (a, b), <i>Φ_st</i> (c, d), and Tajima's <i>D</i> (e, f) simulated for a 22-locus dataset (or a 16-locus dataset in the selection model) with 1,000 replicates; horizontal lines indicate the 95% confidence limits. Lightly shaded squares mark the values for empirical data that fell within the 95% confidence intervals; dark shading indicates empirical values that fell within the 2.5% tails of the posterior predictive distributions.