Compute Posterior Log Randomized Predictive p-values (Log-RPP)

Calculates the posterior log randomized predictive p-values (log-RPP) for a set of observations given the log-PMF and log-CDF matrices from posterior predictive samples.

Usage

post_logrpp(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 randomized posterior predictive p-values for each observation.

Details

The function computes a randomized posterior predictive p-values as follows:

1. A uniform random number is generated for each observation to randomize the probability for zero counts.

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

Examples

NULL
#> NULL