TABLE 3

Performance of τ and the χ2-approximation to the likelihood ratio

Geneenvgagenvagag
SubtypeAABB
Actual Nτ1.2 × 1058.3 × 1052.1 × 10317.8
Mean τ1.96 × 1053.1 × 1067.8 × 10372.7
% cases actual Nτ outside confidence limits5341

The mean estimated Nτ for sequence alignments simulated under known Nτ's, together with the number of times the likelihood of the known Nτ was significantly worse than the estimated (maximum-likelihood) value. The four known Nτ's used correspond to the values estimated for HIV-1 subtype A and B, gag and env genes, when the doubling time was set to 1 yr. In each case the number of simulated sequence alignments used was 100, and the number of sequences in each alignment corresponded to the number of sequences used for the original estimate.

  • a For the subtype B env alignment 25 sequences were simulated in each alignment rather than the 54 used originally to estimate Nτ, to save computational time.