Tables describing model selection for each response variable.

Support for predictors was tested using likelihood-ratio tests assuming χ2 distribution of deviance changes. Order of elimination of nonsignificant predictors is shown in corresponding column. Predictors included in the minimal adequate model (either as a main effect or through interaction with other variables) are in bold face. S2A Table. Model selection process for body mass, tarsus length, relative tarsus length, relative bill width and length. S2B Table. Model selection process for H/L ratio. S2C Table. Model selection process for proportion of immature erythrocytes. S2D Table. Model selection process for complement activity.
(XLSX)