Systematics/Phytogeography / Taxonomie/ Section
Freudenstein, John , Davis, Jerrold I. .
Going out on a limb -- an analysis of resampling branch support.
Resampling branch support analyses (jackknife, bootstrap) are the most widely employed methods of assessing support. Used in a non-statistical sense, these approaches give a dataset-dependent measure of relative branch support for individual clades. Given that formalization of taxonomic groups and conclusions about evolutionary patterns are often made based on perceived strength of clade support, it is important to understand how the details of a resampling analysis affect the results. A fast analysis, in which no branch swapping or sophisticated heuristics are performed, yields unrepresentatively low values. Increased search effort achieves higher values for most clades that fairly rapidly plateau, such that additional effort gives little increase. Beyond this point increased search effort causes some values to decrease. This effect is due in part to the number of trees that are saved in each support replicate and in part to more difficult to search clades being recovered. The frequency-within-replicates approach to summarizing the results of each support replicate results in several percentage points greater values than the strict consensus approach. The sampling with replacement method of the bootstrap, yielding resampled datasets that comprise unobserved combinations of characters, often gives different results than the jackknife, which utilizes subsets of the real dataset and is here viewed as more appropriate.
Log in to add this item to your schedule
1 - Ohio State University, Department of Evolution, Ecology, and Organismal Biology, University Herbarium, 1315 Kinnear Rd., Columbus, Ohio, 43212, USA
2 - Cornell University, L.H. Bailey Hortorium, Department of Plant Biology, Ithaca, New York, 14853, USA
Presentation Type: Oral Paper:Papers for Sections
Location: Room 1/Woodward
Date: Monday, July 28th, 2008
Time: 3:00 PM