Second-order Akaike’s information criterion (AIC_c) comparison of statistical models.

Model selection based on AICc, with a penalty of 2 per added parameter [29]. Models are ordered by adequacy, starting with the minimum adequate model. Model 1.2 is competitive with model 1.1. Model 2.1 and 3.1 do not have competitors. All models are linear mixed models with a Gaussian error structure, and contain bird ID as a random effect. Models 2.1 to 2.5 contain a variance structure based on prey species. a In model 1.1 to 1.5, factor “diet” refers to the diet outside the experimental trials, being either soft or hard-shelled. Factor “group” refers to the order of these diet treatments (group 1 or group 2). In models 2.1 to 2.5, factor “gizzard” refers to gizzard size during the trial, which was either small or large; “species” refers to the prey species being offered, which was either Dosinia or Loripes. In models 3.1 to 3.5 log(gizzard) is a continuous variable that refers to the logarithm of estimated gizzard mass during the trial; species refers to prey species, which was either Dosinia isocardia, Cerastoderma edule or Macoma balthica. The symbol × means that the main terms as well as their interaction are fixed effects in the model. Models 1.4, 2.4 and 3.5 contain only an intercept, no fixed effects. b The number of parameters in the model. c Log likelihood. d Dry ballast mass. Second-order Akaike’s information criterion (AICc) comparison of statistical models.