LBS3.4 Friday, Jan. 4 Common biases in empirical tests of diversification using phylogenetic trees BROCK, Chad D*; ALFARO, Mike E; HARMON, Luke J; Washington State University, Pullman, WA; Washington State University, Pullman, WA; University of Idaho, Moscow, ID, firstname.lastname@example.org
The recent growth in both the number of dated molecular phylogenies and statistical methods for evaluating rates and patterns of cladogenesis has allowed detailed investigations of macroevolutionary patterns by neontologists. Despite the popularity of phylogenetic diversification statistics, little is known about the behavior of many of these analyses, particularly if their assumptions are violated. One key issue, non-random taxon sampling, is likely to be a common problem in empirical data sets. We use tree simulations to investigate the behavior of Pybus and Harvey’s (2000) gamma statistic in situations where taxon sampling is incomplete and non-random. We find that non-random sampling substantially increases the type-1 error rate of the gamma statistic and we present a conservative method for correcting this bias. We also investigate the behavior of the probability calculation of Magallon and Sanderson (2001) for species richness given different methods of choosing clades of interest. We find a significantly elevated type-1 error rate in this approach when individual clades were chosen post-hoc based on their apparent diversity and we present a way to correct for this when estimating p-values. Alternative methods for investigating patterns of diversification across clades and trees are also discussed. Support for this work was provided by an NSF IGERT Fellowship (NSF BCS-0549425) to CDB and NSF DEB 0445453 to MEA.