Weighting improves group support in morphological data sets

GOLOBOFF, PABLO

Fagernes, Noruega

Congreso; XXIV Annual Meeting of the Willi Hennig Society (ver http://folk.uio.no/victoraa/scheduleHennig24.pdf); 2005

Willi Hennig Society

When analyzing a very large rbcL data set, Källersjö et al. (1999) discovered that eliminating third positions (more homoplastic), group support was decreased.  This result contrasted with earlier comparisons (by Goloboff, 1997), on smaller, morphological data sets, where weighting against homoplasy improved group support.  Källersjö et al. interpreted their result as indicating problems with weighting functions that weight against homoplasy, but their result has often been interpreted as a definitive argument against character weighting in general.  The problem is revisited here.  In real-sized morphological data sets (50-200 taxa), it is shown that weighting consistently improves measures of support and stability.  This poses two alternatives (both of which are discussed).  One is considering that the decrease in support is to be interpreted as evidence against weighting in large molecular data sets (and then the increase in support must be interpreted as an argument in favor of weighting in the case of morphological data sets).  The second alternative is considering that the changes in group support cannot be interpreted as evidence for/against a weighting method.  Any of these alternatives invalidates one of the most often voiced arguments against character weighting, in the case of morphological data sets.  Additionally, it is shown that using properly scaled functions which weight against homoplasy, group support in molecular data sets (with 400-900 taxa) also increases.