Table 2.
Source of uncertainty | Description | Method of propagation |
---|---|---|
Fixed effects | Uncertainty in the fixed effects for each sub-model | 1000 multivariate (or univariate depending on the model definition) normal samples using the means and vcov matrix of the fixed effects |
Phylogenetic random effect | Uncertainty in the amount of evolutionary change in the response variable (e.g. bird titer) that has occurred over each branch of the phylogeny | 1000 multivariate (or univariate for models with a single species-level random effect) normal samples for each branch, with means equal to the conditional modes of the species-level random effect for each branch multiplied by the branch lengths and variance equal to the variance of the conditional modes of the random effects for each branch multiplied by the squared branch lengths |
Phylogenetic tip variation | Evolutionary change that has occurred after the divergence of the species whose response is being imputed from its most closely related species that has an empirically measured (and estimated) response | 1000 multivariate (or univariate for models with a single species-level random effect) normal samples with mean 0 (because of the assumption of Brownian motion), and SD equal to the SD of the species-level random effect multiplied by the length of the final (most recent in time) branch leading to the species in question |
Other random effects | Uncertainty due to variation among studies and infection experiments | 1000 univariate normal samples for each random effect with mean equal to 0 and SD equal to the estimated SD |
Stan model overall uncertainty | Summary of the entire uncertainty associated with the three Stan models used in the transmission steps between mosquitoes and birds (bird-to-mosquito transmission probability, mosquito-to-bird transmission probability, and mosquito biting preference) | 1000 samples from the posterior distributions for each of the Stan models |
Abbreviation: SD, standard deviation; vcov, variance-covariance