Systematics Section / ASPT
Albert, Victor A. .
Pasimony analysis of correlation: a general framework for hierarchical analysis of associations.
There are several topics in the phylogeny-based study of associations that can be generalized under a single framework. one topic refers to previous approaches that have employed completed trees for separate associates in their analyses of correlation. Examples include reconsructing histories of hosts and their parasites, or taxa and their areas of geographic distribution. When most-parsimonious (or other) trees are reconciled ex post facto, primary evidence for putative homology is not being tested for overall congruence. A method is proposed that takes all character evidence from associates into account simultaneously: instead of using organisms as terminals for separate phylogeny reconstructions, associations can be recognized as terminal units, and character data from 2 or more organisms, or organisms and attributes (such as distribution), can be analyzed together. This entails generating a secondary data matrix of associates, the data for which is comprised of concatenated characters specific to particular associations. Inapplicable character data (where no associate/s exist for certain organismal/attribute groups) can be analytically voided using a new algorithm developed by Jan De Laet. Another subject that can be addressed using most-parsimonious congruence testing is the degree of hierarchical association between any data that could reasonably be hypothesized as hierarchically correlated. For example, floral fragrance compounds or ethnobotanically-investigated uses for plants might be related hierarchically by their associated plant species as characters, and as such, provide evidence toward reconstructing biosynthetic pathways or the anthropological evolution of plant uses. Both data-concatenated and simple, albeit heretical, parsimony analysis of association can be evaluted for the strength of hierarchy using resampling techniques such as the jackknife. Where jackknife support exists for a hierarchical relationship among data, then it can be accepted; where it does not, either homplasy or truly non-hierchical relationships may exist. Examples attempted over the past decade demonstrate merit for this framework.
1 - University of Oslo, Botanical Garden, Natural History Museums and Botanical Garden, P.O. Box 1172 Blindern, Oslo, NO-0318, Norway
Presentation Type: Paper
Location: Cottonwood B (Snowbird Center)
Date: Wednesday, August 4th, 2004
Time: 8:30 AM