Table 3.
Bayes Factors (BF) between exponential (Exp) and logistic (Log) growth demographic models for the HIV-1 CRF12_BF and CRF38_BF pol data sets.
| Subtype | Model comparison | ln (BF)a | Evidence against H0b |
|---|---|---|---|
| CRF12_BF | Exp (H0) vs Log (H1) Strict clock | 111.7 (0.6) | Decisive |
| Exp (H0) vs Log (H1) Relaxed clock | 142.0 (0.7) | Decisive | |
| Log Strict (H0) vs Relaxed (H1) clock | 48.0 (0.5) | Decisive | |
| CRF38_BF | Exp (H0) vs Log (H1) Strict clock | 6.9 (0.3) | Decisive |
| Exp (H0) vs Log (H1) Relaxed clock | 10.5 (0.4) | Decisive | |
| Log Strict (H0) vs Relaxed (H1) clock | 8.7 (0.3) | Decisive | |
a ln (BF) = Bayes Factor is the difference (in ln space) of the marginal likelihood of null (H0) and alternative (H1) model. The SE of the estimates is given in parenthesis and was estimated using 1000 bootstrap replicates. b Evidence against H0 was assessed in the following way: ln (BF) < 0 indicates no evidence against the null model; ln (BF) between 0-2.3 indicates weak evidence against the null model, ln (BF) between 2.3-3.4 indicates strong evidence against the null model; ln (BF) between 3.4-4.6 indicates very strong evidence against the null model; and ln (BF) > 4.6 indicates decisive evidence against the null model.