Model selection results for candidate general linear mixed models used in the evaluation of the relationship between δ15N and δ13C values of western sandpiper blood (the response), bill length, and sex.

Site was included as a random effect: all δ15N models included a random intercept and slope (by bill length for each site) effect, while δ13C models included only a random intercept. The number of parameters (K) includes a parameter for the intercept, variance of the random intercept, the residual variance, and also for the variance of the random slope, and the correlation between the estimated variances of the random components (if included in the model). Deviance is equal to -2 x log-likelihood and was used to calculate AICc (Akaike’s Information Criterion corrected for small sample size). Competing models were ranked according to ΔAIC­c and Akaike weight (AICcw). The number of observations used for all models was 122.