Premium
A biometric study of the effects of growth on the analysis of geographic variation: Tooth number in Green geckos (Reptilia: Phelsuma )
Author(s) -
Thorpe R. S.
Publication year - 1983
Publication title -
journal of zoology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.915
H-Index - 96
eISSN - 1469-7998
pISSN - 0952-8369
DOI - 10.1111/j.1469-7998.1983.tb04258.x
Subject(s) - biology , principal component analysis , bivariate analysis , ordination , covariance , morphometrics , variation (astronomy) , geographic variation , biometrics , population , statistics , ecology , evolutionary biology , demography , mathematics , physics , computer security , sociology , astrophysics , computer science
Using tooth number in Green geckos from the Seychelles it is shown that growth influences can perturb an analysis of geographic variation in two ways. First, the pervasive influence of growth results in a high within‐group correlation between characters as each character is repeatedly measuring similar aspects of the phenotype. Second, there can be a bias in the growth stage of the sample such that one population is represented mainly by juveniles whilst another is represented mainly by adults. The former source of perturbation (correlation) is shown to be more influential than the latter (bias). Growth is negated, either by bivariate regression using the pooled‐within group slope, or by multiple‐group principal component analysis. The relative advantages of these procedures is discussed but in practice they gave extremely similar results. When growth is negated in this way the choice of final ordination technique is unimportant. When growth is not negated then the choice of technique is extremely important. Techniques that take account of the within‐group covariance (canonical variates) give results that are similar to the growth‐free analyses even when growth is not negated. Several facets of morphometrics such as logarithmic transformation, assessing the contribution of characters and pooling irrespective of group are critically discussed.