Candidate models using body mass^0.75.

<p>Models in bold were used for coefficient averaging.</p><p>The corrected Akaike information criterion (AICc) was used to rank the models using the difference with the lowest observed AICc (e.g., ΔAICc).</p><p>Akaike weights represent the probability that a given model best reduced the information loss for the observed data.</p><p>T, Ambient temperature; P Total body lipid (%); L, Location; A, Age; S, Sex.