Statistical analysis
All analyses were conducted in R 4.1.0 (R Core Team 2013) and
significance was based on log-likelihood ratio tests. To test if local
pathogens are more deadly than novel pathogens, we conducted a
Cox-proportional hazards survival analysis in which the response was
mortality and the predictors were the additive terms: distance, host
species, Bd strain, and host mass (survival package) (Therneau 2014).
Distance was a continuous variable defined as the log-linear Euclidean
distance between the collection location of the host and Bd strain
(log10(km)) (Johnson et al. 2021). We conducted a generalized
linear model with a normal error distribution to determine the effect of
the same predictors on Bd zoospore load (adjusted for copy number and
log10 transformed) two weeks after exposure or on day of
death if an animal died before day 14 (base package). This GLM included
both infected and uninfected animals and a fixed effect for swab date.
We conducted a GLM with a binomial error distribution to determine the
effect of the same predictors on Bd prevalence (stats package). In the
Cox regression and both GLM models, we controlled for potential effects
of host mass (g) as a covariate.