Parrot data
As reported by McElroy et al. (2018), the parrot data set contained considerable amounts of adapter sequences, ranging from 4 to 41% of the data, which we excluded in order to generate more accurate abundance estimates. Our de novo assembly of the parrot mitochondrial genome, generated from sample SRR6214434, was 19,278 bp in length. The per-sample mapping rates varied over two orders of magnitude, from 0.02 to 1.9%. A PCA run on individual mitochondrial alleles separated the samples into two groups corresponding to the phylogroups inferred by McElroy et al. (2018). The intercept estimate of the nuclear proportion of mitochondrial inserts was 0.0043% (95% CI: 0.0026%-0.070%) The mapping depth estimate was 0.017%. As with the grasshopper data, we analysed the minor allele frequencies at each site that showed fixed mitochondrial difference between the populations (Figure 4). This resulted in a diverged sites estimate of 0.0033% (SE 0.00016%).