krige (version 0.6.2)

heidel.welch: Heidelberger and Welch Diagnostic for MCMC

Description

Conducts a Heidelberger and Welch convergence diagnostic on MCMC iterations.

Usage

heidel.welch(object, pvalue)
# S3 method for krige
heidel.welch(object, pvalue = 0.05)
# S3 method for summary.krige
heidel.welch(object, pvalue = 0.05)
# S3 method for default
heidel.welch(object, pvalue = 0.05)

Arguments

object

An matrix or krige/summary.krige object for which a Heidelberger and Welch diagnostic is desired

pvalue

Alpha level for significance tests. Defaults to 0.05.

Value

A matrix in which the first row consists of the values of the Cramer-von Mises test statistic for each parameter, and the second row consists of the corresponding p-values. Each column of the matrix represents another parameter of interest. A significant result serves as evidence of nonconvergence, so non-significant results are desired.

Details

This is a generic function currently works with matrix, krige, and summary.krige objects. It is a simplified version of the Heidelberger and Welch test for use with this package.

This is an adaptation of a function in Plummer et al.'s coda package. Heidelberger and Welch's (1993) test for nonconvergence. This version of the diagnostic only reports a Cramer-von Mises test and its corresponding p-value to determine if the chain is weakly stationary with comparisons of early portions of the chain to the end of the chain.

References

Philip Heidelberger and Peter D. Welch. 1993. "Simulation Run Length Control in the Presence of an Initial Transient." Operations Research 31:1109-1144.

Martyn Plummer, Nicky Best, Kate Cowles and Karen Vines. 2006. "CODA: Convergence Diagnosis and Output Analysis for MCMC." R News 6:7-11.

See Also

heidel.welch.krige, heidel.welch.summary.krige, geweke

Examples

# NOT RUN {
# Load Data
data(ContrivedData)

# Set seed
set.seed(1241060320)

M <- 100

contrived.run <- metropolis.krige(y ~ x.1 + x.2, coords = c("s.1","s.2"), 
  data = ContrivedData, n.iter = M, n.burnin = 20, range.tol = 0.05)

heidel.welch(contrived.run)
heidel.welch(summary(contrived.run))
heidel.welch(contrived.run$mcmc.mat)
# }
# NOT RUN {
# }