Plot variance partitions

Plot variance partitions of the different model components for occupancy and detection submodels.

Usage

plot_partitions(fit, scales = FALSE, ...)

Arguments

fit

A fitted model object from fit_model().

scales

logical. If FALSE (default), plots variance simplex partitions \(\boldsymbol{\phi}\). If TRUE, produces component scales by plotting \(\sqrt{W \cdot \boldsymbol{\phi}}\), where \(W\) are variances of linear predictors. Useful for sparse simplexes, where few components account for most of the variance.

...

Additional arguments passed to ggdist::stat_pointinterval().

Value

A ggplot object with occARU-specific attributes attached:

plot_data

The tibble used to produce the plot.

Details

The occARU model uses global-local shrinkage priors for the occupancy and detection submodels, where half-Student-t priors are used for the variances of both linear predictors which are simplex partitioned via either Dirichlet or logistic-normal decomposition. Variance decomposition only occurs when there is more than one model component. Partitions exist for species-level intercepts, and species-level slopes, sites effects, survey effects, and Poisson OLREs (with one mean partition and one for species-level deviations). The species-level scales for site and survey effects and OLREs are produced by additional simplex decomposition of the species-level components.

See also

fit_model() plot_intercepts(), plot_coefficients(), plot_sites(), plot_surveys(), plot_correlations(), plot_realised()