Character Transformations in Phenetic Studies Using Continuous Morphometric Variables

Publisher Website
Google Scholar
Abstract
The stability (measured using matrix correlations) of phenetic classifications based on partitions of continuous morphometric data (skeletal measurements from 3 groups of birs) was compared for analyses involving 2 transformations: removal of size and removal of the common part (an estimate of the taxon of the study group). Linear regression was used in both transformations to remove that portion of the variance accounted for by the estimate of either size or the common part. The size estimate was a composite variable computed as the mean of 3 bone lengths. The estimate of the common part was the set of measurements used to form the data set to be analyzed but taken on a similar species outside the study group. Results derived from 3 similarity measures (product-moment correlations, average taxonomic distances and Manhattan distances) were also compared. In all cases distance measures produced significantly higher average congruence between classifications than did correlations. For 2 data sets the transformation to remove the common part resulted in significantly higher mean congruence (for distances) than did the transformation to remove size, for the 3rd data set there was no significant difference. The common-part-removed transformation is suggested for use in phenetic studies of continuous morphometric-data.

This publication has 3 references indexed in Scilit: