Comparison of linear and quadratic models for log-log regressions of size-energy relationships, examined with both the indirect and direct methods and both the population and size-class approaches.

Best models were selected on the basis of the small sample size corrected Akaike's Information Criterion (AICc); Δi is the difference in AICc between the best and the alternate model, and wi is the Akaike weight, the weight of evidence for each model given the data (where 1.000 represents the highest likelihood of the model relative to the alternate model).