TABLE 5.
Simulation results for the method with Bayesian variable selection
| True: | Balancing selection
|
Neutrality
|
Directional selection
|
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Classified: | Balancing | Neutrality | Directional | Balancing | Neutrality | Directional | Balancing | Neutrality | Directional | Specificity | Power | |
| s100 | 9 | 41 | 0 | 8 | 890 | 2 | 0 | 0 | 50 | 0.99 | 0.59 | |
| s050 | 12 | 38 | 0 | 7 | 889 | 4 | 0 | 0 | 50 | 0.99 | 0.62 | |
| s020 | 7 | 42 | 1 | 5 | 893 | 2 | 0 | 0 | 50 | 0.99 | 0.57 | |
| s100-Fb | 13 | 37 | 0 | 8 | 889 | 3 | 0 | 0 | 50 | 0.99 | 0.63 | |
| s050-Fb | 10 | 39 | 1 | 5 | 894 | 1 | 0 | 0 | 50 | 0.99 | 0.60 | |
| s020-Fb | 3 | 47 | 0 | 8 | 888 | 4 | 0 | 0 | 50 | 0.99 | 0.53 | |
| s100-Fb-40 | 6 | 43 | 1 | 15 | 883 | 2 | 0 | 0 | 50 | 0.98 | 0.56 | |
| Neutral | 1 | 999 | 0 | 1.00 | ||||||||
Numbers of loci simulated under balancing selection, neutrality, and directional selection that were classified in each category by the reparameterized Bayesian regression analysis including variable selection are shown. A locus was classified as subject to selection if P(δi = 1 | data) ≥ 0.17 and otherwise was classified as neutral. For nonneutral loci the corresponding Fst-value was used to decide whether the locus is subject to directional or balancing selection.