Open AccessIs logarithmic transformation necessary in allometry?
Author(s) -
Packard Gary C.
Publication year - 2013
Publication title -
biological journal of the linnean society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.906
H-Index - 112
eISSN - 1095-8312
pISSN - 0024-4066
DOI - 10.1111/bij.12038
Subject(s) - allometry , logarithm , bivariate analysis , transformation (genetics) , logarithmic scale , field (mathematics) , bivariate data , statistics , biology , exploratory data analysis , scale (ratio) , variation (astronomy) , econometrics , ecology , mathematics , computer science , mathematical analysis , gene , biochemistry , physics , quantum mechanics , acoustics , astrophysics , pure mathematics
The M etabolic T heory of E cology ( MTE ) transformed the field of biological allometry from a discipline that is focused on description to a discipline that is focused more on formulating and testing theory. However, much of the empirical research providing essential background for the MTE – as well as research to test predictions of the theory – is based on the ‘allometric method’, which is a simple procedure for estimating the parameters in a two‐parameter power function y = a x bby exponentiating the equation for a straight line fitted to logarithmic transformations of the original bivariate data. The allometric method has been in widespread use for so long that many investigators now apply the procedure mechanically and without due consideration for limitations of the approach. What has been missing from much of the contemporary research on allometric variation is exploratory analysis of untransformed data and graphical validation of the fitted model. I use two examples from the current literature: (1) to demonstrate the utility of exploratory analysis; (2) to illustrate how transformation may lead investigators to conclusions that are not supported by their data; and (3) to show how nonlinear regression may obviate the putative need to transform. The MTE (and other theories pertaining to patterns of allometric variation) will benefit from greater awareness that the traditional allometric method is not well suited for fitting statistical models to data expressed in the arithmetic scale. © 2013 The Linnean Society of London, Biological Journal of the Linnean Society , 2013 , 109 , 476 – 486 .