The effects of different numerical techniques on the phenetic classification of bees of the Hoplitis complex (Megachilidae)

SUMMARY The characters of the data matrix of the Hoplitis complex of bees were standardized and similarities computed as distance and correlation coefficients. These similarity matrices were clustered by a variety of techniques. Standardization of character states reduces the isolation of unusual OTU's. The classification based on standardized characters agreed well with one prepared by classical taxonomic methods. Classifications resulting from distance coefficients vary more with the clustering procedure than do classifications based on correlation coefficients. Differences in clustering procedures affect the resulting classification. Differential weighting of stems during clustering caused distinct effects, especially with matrices of distance coefficients and to a lesser extent when Spearman's sums of variables method was used. Only minor differences exist between results of pair-group and variable-group clustering methods. The implications of these findings for numerical taxonomy are discussed.