Table 2.
Simulation | LRT false positive rate | median ω+ and p+ MLEs | ||
---|---|---|---|---|
M1a - M2a | G1aDT - G2aDT | M2a | G2aDT | |
1a (simple, 5 taxa) | 0.49 | 0.04 | ω+ = 6.08 | ω+ = 1.16 |
p+ = 0.37 | p+ = 0.33 | |||
1b (complex, 5 taxa) | 0.22 | 0.04 | ω+ = 10.9 | ω+ = 1.37 |
p+ = 0.28 | p+ = 0.20 | |||
1c (complex, 10 taxa) | 0.48 | 0.04* | ω+ = 1.35 | ω+ = 1.02 |
p+ = 0.49 | p+ = 0.35 |
One hundred replicates (sequence length = 300 codons) were simulated for each scenario. Simulation 1a is based on a simple model (equal DNA exchangeabilities and equal codon frequencies) evolved over a 5-taxon tree. Simulation 1b is based on a more complex generating process using DNA exchangeabilities and codon frequencies derived from a real dataset. Simulation 1b was extended to the case of a 10-taxon tree. Codon models fitted to simulation 1a assumed equal codon frequencies (fequal), and those fitted to simulation 1b used GY94-style F3 × 4 codon frequencies. The asterisk symbol (*) indicates that the results for simulation 1c under the 10-taxon tree is based on 97 replicates due to convergence problems with some datasets