Statistical approaches to handle proportions as characters
Wiesemüller, Berhard;Rothe, Hartmut
The handling of morphometric proportions as characters in phylogenetics is methodically hardly elaborated. Common approaches are either to search for statistically significant differences in"index"-values, usually quotients of measures, or to compare regression lines for logarithmic values. The former have their weakness in an insufficient consideration of allometric proportions, the latter in a lack of appropriate methods for testing statistical significance of differences. This contribution presents a methodical approach which considers both allometry and statistical significance. Moreover, not only significance, but also the size of mean differences, as compared to standard deviations, is calculated. If differences are too small, they do not allow to draw a clear distinction between compared groups, and therefore have a low quality as characters. At the beginning of the procedure a test for gradient differences between allometry lines is applied, and if no such are found, a second test for parallel shift between the two lines. If a difference is found, the extent of which exceeds a prearranged limit, it is accepted as a character difference suitable for phylogenetic reconstruction. This method was practically applied by Wiesemüller & Rothe (1999) in a phylogenetic investigation of New World monkeys (Platyrrhini,Primates).