R_hat: Compute Gelman-Rubin statistic

In loelschlaeger/RprobitB: Bayesian Probit Choice Modeling

Compute Gelman-Rubin statistic

Description

This function computes the Gelman-Rubin statistic R_hat.

Usage

R_hat(samples, parts = 2)

Arguments

samples	A vector or a matrix of samples from a Markov chain, e.g. Gibbs samples. If samples is a matrix, each column gives the samples for a separate run.
parts	The number of parts to divide each chain into sub-chains.

Value

A numeric value, the Gelman-Rubin statistic.

References

https://bookdown.org/rdpeng/advstatcomp/monitoring-convergence.html

Examples

no_chains <- 2
length_chains <- 1e3
samples <- matrix(NA_real_, length_chains, no_chains)
samples[1, ] <- 1
Gamma <- matrix(c(0.8, 0.1, 0.2, 0.9), 2, 2)
for (c in 1:no_chains) {
  for (t in 2:length_chains) {
    samples[t, c] <- sample(1:2, 1, prob = Gamma[samples[t - 1, c], ])
  }
}
R_hat(samples)

loelschlaeger/RprobitB documentation built on Oct. 15, 2024, 11:08 a.m.