Table 1.
Performance summaries of qpAdm rotation analysis for the four demographic models and different performance metrics.
| Plausibility criteria | Model 1 FPR | Model 1 FDR | Model 1 QTP | Model 1 QTP-binary | Model 2 FPR | Model 2 FDR | Model 2 QTP | Model 2 QTP-binary | Model 3 FPR | Model 3 FDR | Model 3 QTP | Model 3 QTP-binary | Model 4 FPR | Model 4 FDR | Model 4 QTP | Model 4 QTP-binary |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| P-value 0.01 | 0.3178 | 0.8443 | 0.6439 | 0.0000 | 0.3014 | 0.8063 | 0.6585 | 0.0031 | 0.3298 | 0.8417 | 0.6389 | 0.0000 | 0.3209 | 0.8387 | 0.6467 | 0.0010 |
| P-value 0.05 | 0.2753 | 0.8407 | 0.5983 | 0.0000 | 0.2549 | 0.7946 | 0.6273 | 0.0086 | 0.2823 | 0.8297 | 0.6238 | 0.0041 | 0.2742 | 0.8293 | 0.6161 | 0.0040 |
| P-value 0.01 + weights [0,1] | 0.1185 | 0.5890 | 0.7611 | 0.1210 | 0.1283 | 0.5921 | 0.7619 | 0.1295 | 0.1179 | 0.5943 | 0.7185 | 0.1251 | 0.1115 | 0.5826 | 0.7241 | 0.1371 |
| P-value 0.05 + weights [0,1] | 0.0983 | 0.5737 | 0.6993 | 0.1479 | 0.1072 | 0.5744 | 0.7122 | 0.1566 | 0.0966 | 0.5658 | 0.6834 | 0.1656 | 0.0918 | 0.5659 | 0.6761 | 0.1553 |
| P-value 0.01 + weights [0,1] ± 2 SE | 0.0743 | 0.6244 | 0.6076 | 0.1170 | 0.0805 | 0.6103 | 0.6396 | 0.1307 | 0.0675 | 0.6444 | 0.5213 | 0.1159 | 0.0653 | 0.6282 | 0.5372 | 0.1275 |
| P-value 0.05 + weights [0,1] ± 2 SE | 0.0601 | 0.6097 | 0.5539 | 0.1426 | 0.0662 | 0.5924 | 0.5966 | 0.1591 | 0.0541 | 0.6135 | 0.4952 | 0.1523 | 0.0523 | 0.6146 | 0.4955 | 0.1421 |
| All single-source models rejected | ||||||||||||||||
| P-value 0.01 + weights [0,1] | 0.0699 | 0.6312 | — | — | 0.0739 | 0.6306 | — | — | 0.0653 | 0.6504 | — | — | 0.0639 | 0.6296 | — | — |
| P-value 0.05 + weights [0,1] | 0.0600 | 0.6087 | — | — | 0.0652 | 0.6051 | — | — | 0.0555 | 0.6111 | — | — | 0.0548 | 0.6057 | — | — |
| P-value 0.01 + weights [0,1] ± 2 SE | 0.0675 | 0.6420 | — | — | 0.0701 | 0.6303 | — | — | 0.0620 | 0.6630 | — | — | 0.0604 | 0.6427 | — | — |
| P-value 0.05 + weights [0,1] ± 2 SE | 0.0565 | 0.6223 | — | — | 0.0605 | 0.6062 | — | — | 0.0512 | 0.6251 | — | — | 0.0503 | 0.6225 | — | — |
| All single-source models rejected and significant f3-statistics | ||||||||||||||||
| P-value 0.01 + weights [0,1] | 0.0524 | 0.6182 | — | — | 0.0559 | 0.6012 | — | — | 0.0530 | 0.6554 | — | — | 0.0514 | 0.6330 | — | — |
| P-value 0.05 + weights [0,1] | 0.0427 | 0.5950 | — | — | 0.0454 | 0.5736 | — | — | 0.0428 | 0.6247 | — | — | 0.0416 | 0.6083 | — | — |
| P-value 0.01 + weights [0,1] ± 2 SE | 0.0523 | 0.6252 | — | — | 0.0554 | 0.6039 | — | — | 0.0525 | 0.6635 | — | — | 0.0506 | 0.6425 | — | — |
| P-value 0.05 + weights [0,1] ± 2 SE | 0.0425 | 0.6030 | — | — | 0.0450 | 0.5764 | — | — | 0.0422 | 0.6341 | — | — | 0.0408 | 0.6208 | — | — |
Each cell contains the average of each performance metric under different qpAdm plausibility criteria. The averages are over the parameter range of the historical period (admixture generations ≤ 100 and median pairwise FST > 0 and ≤0.02). The performance metrics are as follows: FPR, false positive rate; FDR, false discovery rate; QTP, qpAdm test performance; and QTP-binary, qpAdm test performance provided that only the true model fits the data. Each performance metric is evaluated under a different model plausibility criterion for the four demographic models. Their averages are printed in each cell.