Skip to main content
. 2009 May 21;58(1):130–145. doi: 10.1093/sysbio/syp017

FIGURE 5.

FIGURE 5.

The effect of ambiguous characters on Bayesian posterior probabilities under different models of among-site rate variation. Axes and shades of gray have the same meaning as in Figure 2b. Each row corresponds to the top row of Figure 4 (Gene B is effectively invariable). Results for models of rate variation are shown: discrete gamma with 4 rate categories (Γ), invariable sites (I), and partitioned with variable rate prior (P). The subscript f indicates that the rate variation parameter was fixed at the true value, removing the effect of the prior on that parameter. Note that in each case, the light gray lines show the results from analyses of data sets in which both Genes A and B were completely unambiguous (i.e., the control). Under the (incorrect) gamma model of rate heterogeneity, the posterior probabilities were slightly biased even for the unambiguous data sets.