Fig 4. The factors that affect variables related to host pathology at the infection peak.

a- b) SMI: standardised mass index; c- d) temperature; e- f) haematocrit; g- h) oxidative stress: membrane resistance (time taken to haemolyse half of red blood cells when faced with free radical attack). The left hand panels (a, c, e, g) show the standard weight of the predictor parameters affecting the host response variables and the number of candidate models in the 95% confidence interval model set that they occur in. Note that the interaction term indicative of host LA (Infection x Origin x Release site) is not always present. In the right hand panels (b, d, f, h) the LA effect (Infection x Origin x Release site) is explored and in these models, the LA term is explicitly retained. The predicted model averaged parameter estimates ± 1 se of each response variable is plotted on the original scale. Note that parameters that were included as covariates were set as “fixed”, i.e, the SW = 1, and they are therefore not shown in the plots. For full model formulae and results see tables in S1 File and S3 File. For further details of the method see Fig 1. The prediction, under a scenario of parasite adaptation, is that local birds will suffer more and have lower SMI, temperature, haematocrit and membrane resistance.