bootstrap.est: Boostrap estimation function

In Sondre91/STGARCH: STGARCH

Description

This function simulates a dataset considered to be the "truly observed" data, which are then estimated. We imaging we do not know the true parameter, but want to estimate the bias and correct our estimate. We therefore simulate M new datasets using the first estimate as the true parameters and estimate the parameters for these as well. Here we know the true parameter value and can therefore estimate the bias. We then use this bias estimate to correct the original estimate.

Usage

bootstrap.est(true_param, M = 100, burnin = 200, cutoff = 100,
  dim = c(30, 30, 200), ncores = 32, W = neighbourmatrix(dim[1],
  prod(dim[1:2])), n.iterations = 30)

Arguments

true_param	Numeric, Parameter vector, consiting of α_0 and α_1.
M	Integer, number of bootstrap replicas to be made
burnin	Integer, size of temporal burnin
cutoff	Integer, number of spatial points to be "cut off" in each spatial direction in order to make the simulated process non-circular. If cutoff=0, the process is circular.
dim	Numeric vector of length 3, indicating dimension size on form spatial1 x spatial2 x temporal
ncores	Integer, number of parallel processors to be used.
W	Neighbourhood matrix of size prod(dim[1:2])x prod(dim[1:2])
n.iterations	Integer, maximum number of iterations in the maximazation routine.

Value

Vector of the original estimates and the parametric bootstrap bias corrected estimates.

Sondre91/STGARCH documentation built on May 9, 2019, 1:52 p.m.