Recon. method | Simulation | MAPE (%) | Topo. error | HPD width (%) | Gamma statistic | Sister pair analysis |
|---|
BEAST 2 (UCLN) | Unlinked | 6.3–30.9 | 0–14 | 43–29 | − 1.46–1.09 | None |
| | Continuous | 6.7–30.3 | 0–16 | 35–605 | − 1.52–0.20 | None |
| | Punctuated | 8.8–80.4 | 0–8 | 42–154 | − 2.16–3.92 | Negative |
PAML (ACLN) | Unlinked | 11.4–192.1 | Fixed | 62–178 | − 5.32–9.77 | Positive |
| | Continuous | 9.4–188.0 | Fixed | 57–151 | − 5.15–6.25 | Positive |
| | Punctuated | 15.4–438.9 | Fixed | 73–327 | − 8.4–2.99 | Positive |
PAML (UCLN) | Unlinked | 22.7–245.2 | Fixed | 49–263 | − 6.53–7.51 | Positive |
| | Continuous | 26.2–227.5 | Fixed | 49–379 | − 6.246.65 | Positive |
| | Punctuated | 18.1–535.4 | Fixed | 102–301 | − 8.43–4.83 | Positive |
- Summary of reconstructed tree characteristics. Trees simulated under three models: Unlinked, Continuous, and Punctuated. Molecular dates were then reconstructed using three different methods: BEAST 2 (uncorrelated lognormal model, UCLN), PAML (autocorrelated lognormal model, ACLN), and PAML (UCLN). For each reconstructed tree we calculated median absolute % error of node ages (MAPE %), topological error (Topo. Error) by number of true bipartitions not reconstructed, and uncertainty by mean highest posterior density interval width as a percentage of reconstructed node age (HPD width %). We compared the gamma statistics of the reconstructed trees to the simulated trees and used sister pair analysis to test if we can detect the positive correlation between speciation rate in the reconstructed trees (None = no significant correlation)
