texmex (version 1.3)

thinAndBurn: Process Metropolis output from GPD fitting to discard unwanted observations.

Description

Process observations from Metropolis fitting of GPD models, to thin the output and discard observations from burn-in period.

Usage

## S3 method for class 'bgpd':
thinAndBurn(object, burn, thin)

Arguments

object

Object of class 'bgpd' as returned by gpd called with method="simulate".

thin

thin or its reciprocal must be a positive integer. If integer valued, this specifies the frequency of observations from the simulated Markov Chain which will be retained. If specified as a proportion, this is the proportion of values which

burn

The number of observations from the simulated Markov Chain to be discarded as burn-in. Must be a non-negative integer, for no burn-in use burn=0.

Value

• Object of class bgpd. See Value returned by gpd using method = "simulate" for details.

  Note that the original chain is not discarded when this function is called: thinAndBurn can be called recursively on the original object with different values of burn and thin without the this object getting progressively smaller!

See Also

gpd

Examples

x <- rnorm(1000)
  # For the values of burn and thin below, we should do many more iterations.
  # The number of iterations is kept low here due to the run time allowed
  # by CRAN.
  mod <- gpd(x, qu=.7, method="sim", iter=11000)
  mod
  par(mfrow=c(3, 2))
  plot(mod)
  mod1 <- thinAndBurn(mod,burn=1000, thin=5)
  plot(mod1)