Skip to contents

Compute Posterior Log Middle-value Predictive p-values (Log-MPP)

Source: R/post_logmpp.R
post_logmpp.Rd

Calculates the posterior log middle-value (0.5) predictive p-values (log-MPP) for a set of observations given the log-PMF and log-CDF matrices obtained from posterior predictive samples.

Usage

post_logmpp(log_cdf, log_pmf)

Arguments

log_cdf

Numeric matrix of log cumulative probabilities for each observation (posterior samples × observations).

log_pmf

Numeric matrix of log probability mass function values for each observation (posterior samples × observations).

Value

A numeric vector of log posterior mid-value predictive probabilities for each observation.

Details

The function computes a stabilized version of the posterior predictive p-values:

  1. It sums the log-CDF and log-PMF across posterior samples using log-sum-exp for numerical stability.

  2. Probabilities exactly equal to 0 or 1 are replaced with small bounds (\(10^{-5}\) and \(9 \cdot 10^{-5}\)) to avoid numerical issues in further computations.

Examples

NULL
#> NULL