GRstats: Compute Gelman and Rubin's convergence diagnostics from multicore BayesX models.

Description

This function takes a fitted bayesx object estimated with multiple chains/cores and computes the Gelman and Rubin's convergence diagnostic of the model parameters using function gelman.diag provided in package coda.

Usage

GRstats(object, term = NULL, ...)

Value

An object returned from gelman.diag.

Arguments

object: an object of class "bayesx", returned from the model fitting function bayesx using the multiple chain or core option.
term: character or integer. The term for which the diagnostics should be computed, see also function samples.
...: arguments passed to function gelman.diag.

Author

Nikolaus Umlauf, Thomas Kneib, Stefan Lang, Achim Zeileis.

Examples

Run this code

if (FALSE) {
## generate some data
set.seed(111)
n <- 500

## regressors
dat <- data.frame(x = runif(n, -3, 3), z = runif(n, -3, 3),
   w = runif(n, 0, 6), fac = factor(rep(1:10, n/10)))

## response
dat$y <- with(dat, 1.5 + sin(x) + cos(z) * sin(w) +
   c(2.67, 5, 6, 3, 4, 2, 6, 7, 9, 7.5)[fac] + rnorm(n, sd = 0.6))

## estimate model
b <- bayesx(y ~ sx(x) + sx(z, w, bs = "te") + fac,
   data = dat, method = "MCMC", chains = 3)

## obtain Gelman and Rubin's convergence diagnostics
GRstats(b, term = c("sx(x)", "sx(z,w)"))
GRstats(b, term = c("linear-samples", "var-samples"))

## of all parameters
GRstats(b, term = c("sx(x)", "sx(z,w)",
  "linear-samples", "var-samples"))
}