Summary of quantitative analyses. ((a) Summary of coefficients and significance values for linear regressions performed. Quantile regression estimates the conditional median of the response variable, unlike ordinary least squares (OLS) that estimates the approximate conditional mean. Robust regression fittings (objective functions used were Huber and Tukey bisquare) are performed by iterated, re-weighted least-squares analyses. The formulae proposed use the average of the coefficients given by the different regression models (see the electronic supplementary material). (b) Component models (delta AICc < 2) that best predict δ13C diet–bioapatite enrichment 
in herbivorous mammals given current data. Body mass alone is the best predictor of enrichment. Other, more complex models that simultaneously assess more potential influences, do not significantly improve prediction of the enrichment in the AICc analyses. Abbreviations: d.f., degrees of freedom; logLik, maximum log-likelihood; AICc, Akaike information criterion corrected for small sample sizes; delta, difference in AICc between the current and the best model; weight, prior weights used in model fitting.)
