Table 2. Average distance between inferred and true ancestry locations in simulated admixed individuals from POPRES data.
| No. Ancestries | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| Naive model | 443 ± 4 (265) | 880 ± 5 (491) | 898 ± 10 (530) | 880 ± 9 (578) |
| SPAMIX haploid model | 458 ± 4 (273) | 557 ± 4 (334) | 620 ± 7 (392) | 665 ± 7 (449) |
| SPAMIX diploid model | 443 ± 4 (265) | 550 ± 4 (326) | 591 ± 7 (367) | 639 ± 7 (423) |
| SPAMIX (logistic) | 75 ± 1 (41) | 236 ± 5 (131) | 363 ± 6 (215) | 419 ± 6 (247) |
The inferred locations are global for the admixed individuals, which marginalizes over all possible uncertainty on the local genome blocks assignment. Simulations assume four generations in the mixture process. Naive model denotes the extension of SPA that ignores admixture-LD. SPAMIX (logistic) represents simulation results starting from haplotypes generated at a location on a map using a Bernoulli sampling from the logistic gradients (see the section Materials and Methods). Parentheses denote the SD, whereas SEM is computed as SD divided by square of number of simulations in each category. POPRES, Population Reference Sample; SPA, spatial ancestry analysis; LD, linkage disequilibrium.