Figure 1.

(a) Empirical distribution of maximum likelihood estimates of λ for 1000 sets of trait values simulated on the Geospiza phylogeny with 13 taxa transformed with λ = 0.6, using σ = 0.18. Most such datasets yielded a maximum likelihood estimate of 0; the mean estimate is $\stackrel{‒}{\lambda }=0.35$. (b) As above, but simulating trait values on a much larger phylogenetic tree (a single, simulated Yule tree with 281 tips), again transformed with λ = 0.6. The estimated values now cluster around the true value, and have mean $\stackrel{‒}{\lambda }=0.59$. (c) The data can be more informative about some parameters than others: shown is the empirical distribution of maximum likelihood estimates of the diversification rate σ for the same simulations as in (a). The mean of the distribution is $\stackrel{‒}{\sigma }=0.18$, matching the value used in the simulations.