R/prior-density-ordinate.R
prior_density_has_provenance.Rdprior_density_has_provenance() tells whether the prior measure of x is
available from deterministic provenance, so that its density heights,
ordinate values and region probabilities are evaluated by BayesTools (from
structural routes, or from numerical grids refined under error control).
It is the public signal for densities without such provenance, whose
region probabilities a caller has to obtain in another way (e.g. from
prior draws).
prior_density_has_provenance(x)TRUE or FALSE.
A BayesTools prior object is its own provenance. A
prior_linear_density has provenance when its builder recorded the prior
measure it describes (every density built by the BayesTools prior-density
builders, e.g. JAGS_formula_prior_density(), parameter_prior_density()
and the prior densities of marginal_posterior()), unless that measure
contains a multiply_by product without a structural density route. It has
none when it is a numerical density grid without that record (e.g. the
standard-deviation components of nested variance allocations) or such a
product: both are used for plotting only, and their heights and region
probabilities stop. A density of point masses only is exact without a
record.
Provenance does not make an ordinate exact: a combination without a
structural route (e.g. three non-normal terms) has provenance and refined
grid region probabilities, while its ordinates are "unknown", so point
hypotheses need prior_ordinate_status(). The method field of
prior_density_ordinate() does not separate the cases: it is
"unsupported_provenance" for a grid without provenance and for a
combination without a structural route alike, and "named_transform" for
such a combination under an output transformation.