Creates exact posterior measure metadata used to distinguish a continuous posterior from one containing atoms. Supply NULL to explicitly declare that the posterior has no point masses.

posterior_atom_attribute(point_masses = NULL, source = "user")

Arguments

point_masses

optional data frame or list with x and mass entries.

source

short label describing the source of the declaration.

Value

A posterior atom metadata object for the atoms metadata of posterior draws (posterior_metadata(x, "atoms") <- ). Posterior point masses are read only from these metadata: precomputed posterior densities (posterior_density_attribute()) describe the continuous part and carry no point masses. BayesTools producers can additionally declare positional scalar marginals, named in the full draw-column order. Scalar extraction prefers these declarations; a continuous joint measure can have a point marginal, for example the last level of an ordered point-total prior. When a producer can certify only scalar marginals, the object records joint_declared = FALSE and retains its typed joint refusal in joint_unavailable. Its empty joint table is dimensional storage. It does not declare an atom-free joint law. Scalar selection uses only certified marginals; whole-law inference retains the joint refusal.