BayesFactor: Estimate Bayes Factor

In extRemes: Extreme Value Analysis

Estimate Bayes Factor

Description

Estimate Bayes factor between two models for two “fevd” objects.

Usage

BayesFactor(m1, m2, burn.in = 499, FUN = "postmode", 
    method = c("laplace", "harmonic"), verbose = FALSE)

Arguments

m1, m2	objects of class “fevd” giving the two models to be compared.
burn.in	numeric how many of the first several iterations from the MCMC sample to throw away before estimating the Bayes factor.
FUN	function to be used to determine the estimated parameter values from the MCMC sample. With the exception of the default (posterior mode), the function should operate on a matrix and return a vector of length equal to the number of parameters. If “mean” is given, then colMeans is actually used.
method	Estimation method to be used.
verbose	logical, should progress information be printed to the screen (no longer necessary).

Details

Better options for estimating the Bayes factor from an MCMC sample are planned for the future. The current options are perhaps the two most common, but do suffer from major drawbacks. See Kass and Raftery (1995) for a review.

Value

A list object of class “htest” is returned with components:

statistic	The estimated Bayes factor.
method	character string naming which estimation method was used.
data.name	character vector naming the models being compared.

Author(s)

Eric Gilleland

References

Kass, R. E. and Raftery, A. E. (1995) Bayes factors. J American Statistical Association, 90 (430), 773–795.

See Also

fevd

Examples

data(PORTw)
fB <- fevd(TMX1, PORTw, method = "Bayesian", iter = 500)
fB2 <- fevd(TMX1, PORTw, location.fun = ~AOindex,
    method = "Bayesian", iter = 500)

BayesFactor(fB, fB2, burn.in = 100, method = "harmonic")

extRemes documentation built on Nov. 19, 2022, 1:07 a.m.