The performance of GA model selection with in estimating the number and membership of rate classes as well as rate values from simulated data.

measures the simulated pairwise sequence divergence (expected substitutions/nucleotide site); , standard deviation (averaged over replicates) of estimated rates from the generating values; , the proportion of simulations for which the correct number of rate classes are inferred; , the proportion of simulations which are under-fitted, , the proportion of simulations which are over-fitted, and , the mean Rand C-statistic [35] between rate clusters in the generating model and that in the inferred models.