BPfit: Sequential break point copula model.

In LucasDowiak/dynocopula: Implementation of markov switching, smooth transition, and sequential break point copulas

Sequential break point copula model.

Description

BPfit performs a recursive search over a bivariate time series of uniform marginal distributions.

Usage

BPfit(x, y, fam1, fam2 = NULL, parallel = FALSE, date_names = NULL,
  ncores = 2)

Arguments

x	A numeric vector of uniform marginal values.
y	A numeric vector of uniform marginal values.
fam1	An integer representing the family of the copula to use. If fam2 is not NULL, this value indicates the copula family of the first copula component.
fam2	A second (optional) integer indicating the use of a mixture copula and the family of the second copula component. Defaults to NULL.
parallel	Logical switch to run the breakpoint search in parallel.
date_names	Character vector of (optional) date/timestamp names for the marginal distributions.
ncores	Integer specifying the number cores to use in the parallelization. If the user specifies more cores (real or logical) than the CPU can support, the max number of supported cores will be used.

Details

The sequential break point model can be more computationally expensive than MSfit and STfit, especially as the length of the time series grows. This is due to the sequential hunt for break points.

In the model's favor, robust standard errors can be calculated for each respective regime. In addition to the standard errors obtained by inverting the hessian, outer product of gradients (OPG) and sandwich estimates are also available. Standard errors based on the sandwich estimator are used in the summary.

Value

BPfit returns an S3 object of class seqBreakPoint.

The summary, plot, coef, and logLik methods will give a decent snapshot of the model.

An object of class seqBreakPoint is a list of lists, one for each regime. Each individual list contains the following components:

pars	a vector of coefficients for the copula
log.likelihood	log-likelihood value for the regime
dep.measures	a list tail dependence and Kendall's tau
emp_hess	the emprical hessian
opg	the outer product of the gradient
sandwich	the sandwich estimator
family	integers recorded which copula family to used
points	start and ending index values that subset the marginal series for the regime

In addition, the following three attributes are included:

marinal_names	names of the marginal series
initial_bp	information on the initial break points before a re-partition method is applied
class	class of the model

References

Dias, Alexandra and Paul Embrechts, 2004, Change-point analysis for dependence structures in finance and insurance. In G. Szegoe (ed.), Risk Measures for the 21st Century, Wiley Finance Series, 321-335

Dias, Alexandra and Paul Embrechts, 2009, Testing for structural changes in exchange rates dependence beyound linear correlation, European Journal of Finance, 15(7), 619-637

LucasDowiak/dynocopula documentation built on April 12, 2024, 1:32 p.m.