Per-draw gate and point states of a catalog quantity

parameter_gate_states() returns, for each draw, the state of the structural gates and point components that put one catalog quantity on its atoms: the inclusion gates of random-effect variance allocations (allocated SDs and variances, allocation totals, variance proportions, and inclusion indicators), and the component indicator of a mixture or spike-and-slab prior with point components (coefficients, factor levels, and random-effect SDs and variances of such priors). These are the states from which parameter_mixed_posterior() declares the posterior atoms.

parameter_gate_states(fit, selection, draws = NULL)

Arguments

fit

a model fitted with JAGS_fit().

selection

a selection of one catalog quantity, created with parameter_catalog_resolve().

draws

optional draws of the fitted coordinates to read the states from: a numeric matrix with one column per coordinate and one row per draw, an mcmc or mcmc.list object, or a named numeric vector holding one draw (as for JAGS_evaluate_deterministic()). It must contain the gate and indicator coordinates of the quantity. Defaults to the posterior draws of fit.

Value

NULL for a quantity without inclusion gates or point components. Otherwise a list with the per-draw elements

atom

the atom of the quantity in the draw, NA where the quantity is on its continuous part or undefined. When known is FALSE, it holds only the gate atoms.

continuous

whether the quantity is on its continuous part (defined and not on an atom; NA where that is not determined).

defined

whether the quantity is defined in the draw (a variance proportion is undefined without an active component).

event

whether the draw is in the inclusion event of the quantity (see parameter_mixed_posterior()), or NULL when the quantity has no such event.

and the elements

known

a scalar: FALSE when the atom states are not determined by the draws (the component indicator of a prior with point components is not monitored).

prior_atoms

for quantities whose atoms are the point components of a mixture, spike-and-slab, or selection prior, a data frame with the atom locations x and their prior masses mass from the prior's component weights (inclusion probabilities of spike-and-slab priors); NULL for the inclusion gates of variance allocations, whose prior masses are the point masses of parameter_prior_density().

See also

parameter_mixed_posterior(), parameter_catalog()