Skip to contents

Up to a constant that does not depend on the power parameter. The source log likelihoods are shifted to a maximum of zero; the shift multiplies the numerator and Z_S(gamma) alike, so it cancels.

Usage

binomial_power_prior_log_marginal(
  terms,
  power_parameter,
  bilinear = binomial_power_prior_bilinear(terms)
)

Arguments

terms

Output of binomial_power_prior_target_terms().

power_parameter

Power parameters in \([0, 1]\).

Value

One log marginal likelihood per power parameter.