Compute Log Predictive Distributions for a Hurdle Negative Binomial Model of a 'brms' Fit

This function calculates the log probability mass function (log-PMF) and the log cumulative distribution function (log-CDF) for each observation from a fitted hurdle negative binomial model (fitted using brms). The function extracts posterior samples for the model parameters and evaluates the predictive distributions across all posterior draws.

Usage

log_pred_dist_HNB(fit)

Arguments

fit

A fitted brms hurdle negative binomial model object. The model must include the distributional parameters mu, shape, and the hurdle probability zero.

Value

A list with the following components:

lpmf_hat

A matrix of log-PMF values (posterior samples × observations).

lcdf_hat

A matrix of log-CCDF values (posterior samples × observations).

zero_id

Indices of observations with zero counts.

count_id

Indices of observations with positive counts.

Details

For each posterior draw and each observation, the function computes:

• lpmf_hat: Log predictive mass function values using dhurdle.nb().

• lcdf_hat: Log cumulative distribution function values using phurdle.nb() with lower.tail = FALSE.

Examples

if (FALSE) { # \dontrun{
# Example usage:
fit <- brm(bf(y ~ x1 + x2, hu ~ x1), family = hurdle_negbinomial(), data = mydata)
pred_dist <- log_pred_dist_HNB(fit)
str(pred_dist)
} # }