Point-hypothesis eligibility of prior ordinates

prior_ordinate_status() applies the exactness rule of point hypotheses (hypothesis_BF(), Savage_Dickey_BF()) to a prior density at each requested value and reports the outcome instead of stopping. A value is eligible for a Savage-Dickey point hypothesis when its prior ordinate is "regular", classified exactly by prior_density_ordinate(), and has an available log density. Otherwise the row names the condition class and the message with which a point hypothesis at that value stops.

prior_ordinate_status(prior_density, values, labels = NULL)

Arguments

prior_density

a BayesTools prior or a prior_linear_density, as accepted by prior_density_ordinate().

values

finite numeric values of the point hypotheses (at least one).

labels

optional labels of the point hypotheses used in the messages, one per value; defaults to the values.

Value

A data frame with one row per value and the columns

value

the value.

eligible

whether a point hypothesis at the value has an exact regular prior ordinate.

condition

the class of the error a point hypothesis at the value stops with: "BayesTools_point_mass_at_null", "BayesTools_infinite_ordinate", "BayesTools_zero_ordinate", "BayesTools_undefined_ordinate", or "BayesTools_inexact_ordinate" (all also of class BayesTools_hypothesis_ordinate); NA when eligible.

reason

the message of that error; NA when eligible.

continuous_behavior

the behavior of the prior without its point masses at the value ("regular", "zero", "infinite", "undefined", or "unknown"): the ordinate's provenance$continuous_behavior at a point mass and its behavior otherwise.

See also

prior_density_ordinate(), hypothesis_BF()

Examples

spike_and_slab <- prior_mixture(
  list(
    prior("point", list(location = 0), prior_weights = 1),
    prior("normal", list(mean = 0, sd = 1), prior_weights = 1)
  ),
  is_null = c(TRUE, FALSE)
)
prior_ordinate_status(spike_and_slab, c(0, 0.5))
#>   value eligible                     condition
#> 1   0.0    FALSE BayesTools_point_mass_at_null
#> 2   0.5     TRUE                          <NA>
#>                                                                                                               reason
#> 1 There is a point mass in the prior at the exact null hypothesis value. The Savage-Dickey density ratio is invalid.
#> 2                                                                                                               <NA>
#>   continuous_behavior
#> 1             regular
#> 2             regular