Table 2.
Bayes factor (BF) analysis for five species delimitation hypotheses within the S. maltophilia complex, plus genogroups #8 and #10, based on marginal likelihoods computed for each hypothesis by path sampling (see Materials and Methods).
| Species modela | marginal lnL estimate | Model rank | ln-BFb (vs. best) | ln-BFc (vs. previous) |
|---|---|---|---|---|
| split1 | -31347.16515 | 1 (best) | NA | NA |
| lump_Smal+Smc1 | -31384.34206 | 2 | 4.308835∗∗ | 4.308835∗∗ (2 vs. 1) |
| lump_Smal +Smc12 | -31498.71633 | 3 | 5.714071∗∗∗ | 5.432623∗∗∗ (3 vs. 2) |
| lump_Smal +Smc123 | -31521.18269 | 4 | 5.852303∗∗∗ | 3.805166∗∗ (4 vs. 3) |
| lump_Smal +Smc1234 | -31593.24923 | 5 | 6.19882∗∗∗ | 4.970737∗∗ (5 vs. 4) |
aThe best (split1) model assumes that genogroup #10 is a sister clade of S. rhizophila (Figure 1A) and splits the Smc into S. maltophilia (Smal) and four additional species clades (Smc1-Smc4), according to Figure 1B. The following models consecutively lump the S. maltophilia clade with the Smc1-Smc4 clades, with lump_S. maltophilia + Smc1234 representing the whole S. maltophilia complex as a single species (Figures 1A,B). The marginal lnL for the Lump_clades_#8+#10 model is -31465.00904, resulting in a ln-BF = 5.462508∗∗∗ when compared against the split1 (best) model. bThe ln-Bayes factors are computed based on the marginal likelihood estimates ln-BF = log(2(M0-M1)), where M0 is the best model (model 1), which is compared against each of the following ones. ∗indicates positive support [ln(BF) is n the range of 1.1–3]; ∗∗indicates strong support [ln(BF) is n the range of 3–5]; ∗∗∗ indicates overwhelming support [ln(BF) > 5]. cThe ln-Bayes factors are computed as described above, but involve M0 as the model preceding model M1 from the ranked model list. NA, not applicable.