Table 3. Computation times and area under the ROC curves (AUC in percentages) for the analyses of the HsIMM-C, IMM, and SS data sets using the different genome-scan approaches.
| Method | Criterion | Mean (median) computation time, min | HsIMM-C | IMM | SS |
|---|---|---|---|---|---|
| BayeScan | BF | 529 (469) | 60.13 | 53.81 | 62.05 |
| FLK | FLK | 0.16 (0.16) | 58.92 | 61.63 | 62.17 |
| BayEnv2 | XtX | 660 (358) | 70.45 | 61.00 | 72.16 |
| BF | 70.58 | 73.84 | 81.96 | ||
| BayPass (core model) | XtX | 22.6 (22.2) | 61.66 | 61.88 | 65.33 |
| Bfis | 74.36 | 78.91 | 82.29 | ||
| eBPis | 74.33 | 78.78 | 82.22 | ||
| BayPass (STD model) | XtX | 21.4 (17.8)a | 49.85 | 49.16 | 47.72 |
| eBPmc | 74.15 | 78.76 | 82.22 | ||
| BayPass (AUX model) | XtX | 45.3 (44.9)a | 60.60 | 59.82 | 61.08 |
| Bfmc | 58.30 | 65.24 | 70.51 | ||
| BayeScenv | Posterior probability | 510 (478) | 66.93 | 62.34 | 70.36 |
| LFMMb | P-value | 33.0 (30.4)c | 75.58 | 78.29 | 81.98 |
| LFMM–10repb | P-value | 310 (248)c | 76.27 | 79.37 | 82.56 |
Computation times are averaged over the 300 analyses (100 data sets × 3 scenarios).
a
Not accounting for the time required to estimate the covariance matrix (obtained here after running BayPass under the core model).
b
Analyses were carried out using individual genotyping data rather than (population) allele count, which provides the best performance (see, e.g., de Villemereuil et al. 2014).
c
Not accounting for the time required to estimate the number of latent factor K (set here to K = 15).