Evaluation of Some Alternative Procedures Used in Numerical Systematics
1988; Allen Press; Volume: 37; Issue: 1 Linguagem: Inglês
10.2307/2413185
ISSN2330-1198
Autores Tópico(s)Evolution and Paleontology Studies
Resumo-In the context of a numerical systematic analysis of dental occlusal morphology of extant hominoid Primates, consideration of alternative procedures for each of several analytic tasks revealed that: 1) ordination of sample means for hominoid primate molar occlusal landmark co-ordinates gave results similar to those of methodologically more rigorous canonical variate analyses (Hartman, 1986); 2) analysis of metric variables (e.g., group means) standardized by character may give biologically less meaningful results than use of variables not so standardized; 3) nonmetric multidimensional scaling may often give a superior twoor three-dimensional summary of phenetic relationships, compared to ordination techniques restricted to rigid transformation (e.g., principal component or co-ordinate analyses); 4) elimination of a latent vector as an adjustment for intergroup size differences will often be inappropriate; and 5) under certain circumstances regression adjustments for the influence of size on shape (allometry) may be unsuitable. [Numerical systematics; molars; hominoid evolution; stereophotogrammetry.] There are many different procedures available for most analytic operations in numerical systematics. Approaches to data standardization, intergroup size adjustment, ordination, phenetic and cladistic cluster analysis, etc., abound (see Sneath and Sokal, 1973 for one summary); in each case, choice among available options can influence the results and interpretations of a multivariate study. Care must, therefore, be taken to identify procedures most suited to the data and systematic objectives at hand. To aid in making such judgments in conjunction with a multivariate assessment of detailed dental measurements in the living Hominoidea (Hartman, 1986), I empirically tested the efficacy of a variety of approaches to my particular analytic objectives. Because others will be faced with similar choices, I believe implications of these findings to be of sufficiently general interest to warrant detailed description. It should be noted at the outset, however, that some of my results may be conditional 1 Present address: Steve E. Hartman, Department of Anatomy, University of New England, Biddeford, Maine 04005. on the great extent to which variation among taxa exceeds that within for my data. With regard to their role in numerical systematic procedures, the following considerations were addressed: 1) how biologically informative are taxon means alone, compared with a data set that includes within-group variances and covariances?; 2) shouldmetric variables (e.g., group means) be standardized by character before multivariate treatment (as suggested by Sneath and Sokal, 1973:154)?; 3) is there a preference among ordination techniques when the objective is display of taxa in a morphometric space of reduced dimension?; 4) is elimination of a latent vector a suitable adjustment for intergroup size differences?; and 5) under what circumstances are regression adjustments for the influence of size on shape (allometry) of dubious merit? MATERIALS AND METHODS Hominoid species whose dentitions were examined included: 1) the siamang gibbon (Hylobates syndactylus), 2) the Bornean subspecies of the lar group of gibbons (H. lar muelleri), 3) the Bornean orangutan (Pongo pygmaeus pygmaeus), 4) a subspecies of low-
Referência(s)