Skip to contents

Weightfunction Prior

Source: R/priors.R
prior_weightfunction.Rd

Create weightfunction publication-bias priors and their weight-prior helper objects.

Usage

prior_weightfunction(
  side = "one-sided",
  steps = c(0.025, 0.05),
  weights = wf_cumulative(),
  reference = "most_significant",
  prior_weights = 1
)

wf_cumulative(alpha = NULL)

wf_fixed(omega)

wf_independent(prior, scale = "omega")

Arguments

side

character. Either "one-sided" or "two-sided".

steps

numeric vector of p-value cut points. These define length(steps) + 1 p-value bins and must be ordered values in (0, 1).

weights

a weight-prior object created by wf_cumulative(), wf_fixed(), or wf_independent().

reference

character. Reference bin, currently "most_significant".

prior_weights

numeric prior model weight.

alpha

optional positive cumulative-Dirichlet concentration parameters, one per p-value bin. If NULL, prior_weightfunction() uses rep(1, length(steps) + 1). Cumulative weights encode monotone decreasing publication weights relative to the most-significant bin.

omega

fixed publication weights, one per bin; values must be non-missing, nonnegative, and match length(steps) + 1 when used in prior_weightfunction().

prior

continuous simple prior distribution for each non-reference weight. Point, discrete, mixture, and other non-simple priors are invalid.

scale

latent scale for independent weights; either "omega", "log_omega", or the "log" alias. Direct "omega" priors need nonnegative support; "log" is normalized to "log_omega".

Value

prior_weightfunction() returns an object inheriting from prior and prior.weightfunction; the wf_*() helpers return weightfunction_weights helper objects with subclass markers.

Details

Fixed weights must have one value per p-value bin (length(steps) + 1), and the reference bin must have weight 1.

See also

publication_bias_prior_specification

Examples

prior_weightfunction("one-sided", steps = 0.025)
#> omega[one-sided: .025] ~ CumDirichlet(1, 1)
prior_weightfunction(
  side    = "one-sided",
  steps   = c(0.025, 0.5),
  weights = wf_fixed(c(1, 0.8, 0.6))
)
#> omega[one-sided: .025, .5] = (1, 0.8, 0.6)