*2.6. Statistical Analysis*

For the statistical analysis, the relationships between the predictors (transport variant, layer line, flock) and the response variables (baseline CM concentration, CM concentration after transport, returned to baseline value after 72 h, and difference in plumage score before and after transport, transport weight) were analyzed simultaneously using mixed-effects models. The flock was modeled as an unstructured random effect for the model constant (intercept), and the transport variant and the layer line were modeled as ordinary fixed effects. For the continuous response baseline CM concentration, CM concentration after transport, difference in plumage score before and after transport, and transport weight, normal distributions were chosen as observation models. For the binary outcome return to baseline value after 72 h, a logistic regression model was used. Results from this analysis were expressed as odds ratios (OR). For baseline CM concentration, temporal progression was also considered by including time as an unstructured random effect (in contrast to a temporal effect, because of very few unequally distributed time points).

Data were analyzed by using the statistical programming language R [31]. All (generalized) mixed-effects models were estimated by the integrated nested Laplace approximation approach [32] within a fully Bayesian setup.
