TABLE 1.
Prediction ability of the HMM methods and of composite-likelihood methods
| Method
|
|||||
|---|---|---|---|---|---|
| Selection strength | HMMA | HMMB | HMMB-SEG | CLsw | SF |
| Detection power | |||||
| α = 300 | 0.98 | 0.98 | 0.84 | 1.00 | 0.94 |
| α = 500 | 1.00 | 0.99 | 0.94 | 1.00 | 0.98 |
| Average no. of sweep windowsa | |||||
| α = 300 | 1.23 | 1.13 | 1.13 | — | — |
| α = 500 | 1.44 | 1.16 | 1.18 | — | — |
| Average length of the largest sweep window (kb)a | |||||
| α = 300 | 4.52 | 6.19 | 5.97 | — | — |
| α = 500 | 6.20 | 8.64 | 8.32 | — | — |
| Proportion of the largest sweep windows including the selected sitea | |||||
| α = 300 | 0.96 | 0.98 | 0.81 | — | — |
| α = 500 | 0.97 | 0.99 | 0.85 | — | — |
| Average distance from the largest sweep window to the selected site (kb)ab | |||||
| α = 300 | 0.97 | 1.10 | 1.68 | 0.69 | 3.20 |
| α = 500 | 1.15 | 1.45 | 1.92 | 0.90 | 2.40 |
Power to detect a simulated recent selective sweep event (τ = 0.001) is shown. HMMA, three-state model; HMMB, three-state model with estimated background emission probabilities; HMMB-SEG, the same as HMMB but with segregating sites only; CLsw, Kim and Stephan's (2002) method; SF, SweepFinder (Nielsen et al. 2005). n = 30, L = 100 kb. Type I error (percentage of falsely detected sweeps using neutral samples) is 5%. —, irrelevant.
a
Among those replicates where at least one sweep is detected.
b
Computed from the center of the sweep window.