Extract random-effect summary posterior distributions — random_effects_summary

Extracts posterior draws for derived random-effect summary quantities with their canonical prior densities. For allocation summaries, the helper keeps raw Dirichlet allocation weights internal and exposes interpretable scalar summaries such as mean-variance SD-component variance multipliers. Gated total-variance proportions are normalized over active components and omit draws on which every component is excluded because a variance share is undefined there.

Each summary is the parameter_mixed_posterior() of its catalog quantity: its prior density is parameter_prior_density() (for gated allocations the mixed measure with its inclusion-gate atoms), and its posterior atoms are declared from its structure: none when the prior density has no point mass; the inclusion-gate atoms of gated total-variance proportions (at 0 and 1) and allocation totals (at 0), and the point components of a mixture scale prior whose indicator is monitored, with masses from the posterior inclusion- and mixture-indicator draws. Point masses without such a source leave the atom status undeclared, and posterior plots of such summaries stop instead of inferring point masses from the draws.

random_effects_summary_posterior(
  fit,
  summary = c("var_mult", "var_prop", "sd_mult", "sd_total", "var_total", "sd_common",
    "var_common"),
  allocation = NULL,
  component = NULL,
  formula_parameter = NULL,
  simplify_names = FALSE,
  n_prior_points = 4096
)

Arguments

fit: model fit created by JAGS_fit or a derived-draw view from JAGS_with_draws(). Summaries use its supplied draws and local analysis metadata.
summary: semantic quantity to extract: "var_mult", "var_prop", "sd_mult", "sd_total", "var_total", "sd_common", or "var_common". The default "var_mult" exists only for mean-variance SD-component allocations: for a fit with total-variance allocations it stops (request "var_prop" or a total instead).
allocation: optional allocation filter. It matches the stable allocation name given to random_variance_allocation() (the identifier the backend uses), not the public display_name.
component: optional allocation component label filter, matched against the component labels shown in the quantity names.
formula_parameter: optional formula parameter filter.
simplify_names: whether returned quantities should use centrally generated simplified display labels. Defaults to FALSE.
n_prior_points: grid size (n_grid of parameter_prior_density()) of the attached prior densities.

Value

A named list of posterior vectors with class mixed_posteriors and marginal_posterior. The list can be passed to plot_posterior, and the individual vectors carry their prior_density, support, and declared atoms as draw metadata (see posterior_metadata()).