Tree size | Method1 | MAP value | MAP diff2 | Rates diff3 | Times diff4 | Comp time5 |
---|
| | average | average | worst | average | worst | average | worst | relative |
---|
10 leaves | combined | -8046.54 | 0.48 | 3.39 | 0.58 | 1.46 | 0.26 | 0.72 | 1 |
| r × t | -8046.72 | 0.32 | 13.03 | 0.36 | 4.61 | 0.17 | 1.73 | 2 |
|
l
| -8049.34 | 0.012 | 0.20 | 0.00068 | 0.0084 | 0.0015 | 0.05 | 1 |
100 leaves | combined | -65249.12 | 1.61 | 8.60 | 4.72 | 10.64 | 1.24 | 3.39 | 2 |
| r × t | -65257.40 | 4.66 | 23.10 | 14.08 | 21.66 | 5.11 | 9.19 | 10 |
|
l
| -65258.41 | 0.16 | 1.54 | 0.034 | 0.071 | 0.060 | 0.41 | 1 |
- 1The combined method and the r × t-method infer optimal substitution rates and divergence times. The l-method infer optimal branch lengths which are subsequently partitioned into rates and times. The latter problem is much easier and the inferred rates and times less optimal.
- 2Difference in log-likelihood between two MAP runs on the same tree.
- 3Difference in substitution rate norm between two MAP runs on the same tree.
- 4Difference in divergence time norm between two MAP runs on the same tree.
- 5Approximate computation time relative to the l-method.