Methods and theory of phylogenetic inference
Freudenstein, John .
Resampling branch support strategies: balancing accuracy and effort.
Assessment of branch support has become increasingly important as analyses continue to increase numbers of characters, making it impractical to map individual characters on the tree. Jackknife and bootstrap employ resampling strategies to give a dataset-dependent value for each clade. While sometimes put forth as a rapid method to analyze a dataset, the extensiveness of tree search can matter significantly for the result and needs to be balanced with computational effort. Jackknife analysis of an orchid plastid DNA sequence data set comprising 173 terminals and 1180 informative characters was conducted varying three parameters: number of jackknife replicates, numbers of trees held during search (affecting branch swapping), and numbers of search replicates per jackknife replicate. Increasing numbers of replicates reduced the variance for branch support values, such that by 5000 replicates, the standard deviation of support values was less than 1, establishing a basis for the significance of comparisons among treatments. Varying the number of trees saved per search and the number of searches per replicate had similar effect -- increasing search extent resulted in higher jackknife values, with the greatest effect in initial increments and less with subsequent increases, such that an asymptotic pattern was observed. Increased search effort required significantly greater search time than analyses with no branch swapping, but yielded surprisingly greater values in some cases. A point of rapidly diminishing return was reached beyond saving two trees per search or using two searches per jackknife replicate. While jackknifing can provide an efficient way of identifying strongly supported clades, it should not be viewed as a “quick and dirty” method, since realistic support values will only be obtained with sufficient search effort.
1 - Ohio State University, EEOB/ Herbarium, 1315 Kinnear Rd., Columbus, Ohio, 43212, USA
Presentation Type: Symposium
Location: Alpine A (Snowbird Center)
Date: Tuesday, August 3rd, 2004
Time: 11:15 AM