Estimate Diet Proportions Using Multivariate Tweedie Model

Defines predict function that transforms output from a Tweedie Generalized Linear Mixed Model (using 'glmmTMB'), Generalized Additive Model (using 'mgcv'), or spatio-temporal Generalized Linear Mixed Model (using package 'tinyVAST'), and returns predicted proportions (and standard errors) across a grouping variable from an equivalent multivariate-logit Tweedie model. These predicted proportions can then be used for standard plotting and diagnostics. See Thorson et al. 2022 .

mvtweedie

An R package to interpret a Tweedie generalized linear model (GLM) or generalized additive model (GAM) involving multiple classes as an estimate of proportions for each class, implicitly involving a multivariate-logit transformation for parameters and predictions. This approach generalizes the Poisson-to-multinomial transformation to allow for non-integer responses, and can analyze either pre-processed (transformed to proportions) or raw (zero-inflated positive real values) data.

This approach is helpful, e.g., when analyzing diet samples that are heavily zero inflated without pre-processing the samples prior to analysis. In these cases, the Tweedie distribution can be interpreted mechanistically as a thinned and double-marked Poisson point process representing foraging processes.