Figure 2. Variation in plasma antioxidant capacity (z-standardised TAS values) between blackcaps with different levels of multiple status of infection by haemosporidian parasites. Estimates and 95% confidence intervals were computed from a model controlling for uric acid content, residual body mass, year and plasma MDA. The dots are individual estimates, jittered horizontally to reduce overlapping.
Food preference and physiology
We obtained six models with ΔAIC ≤ 2 (Table 2), which were deemed equally good to explain food preference. The model with lowest AICc value included RBC oxidative stress, plasma-MDA (log-transformed), uric acid and year. Given that the best model was not strongly weighted against competitor models, we used a full model averaging of the six models to obtain parameter estimates following (Symonds and Moussalli 2011). Plasma-MDA was the only physiological variable with a significant effect on food preference: individuals with lower log-transformed plasma-MDA preferred fat-enriched food (full model estimate ± adjusted SE = -2.08 ± 1.04, P = 0.047) controlling for an effect of the year (estimate ± adjusted SE = -1.99 ± 1.01, P = 0.049, all other effects with P > 0.07).
Table 2. Selection of the best models to explain food preference as a function of physiological parameters involved in oxidative status. The models are arranged according to the Akaike Information Criterion corrected by sample size (AICc), where the best model (number 1) is the one with lowest AICc. Models with ΔAICc ≤ 2 are shown, indicating their AICc value, the increase in AICc compared to the best model, the model weight and the number of parameters. The same information is shown for the null model (ranked 95) at the end of the table.