Title: | Change Point Tests for Joint Distributions and Copulas |
---|---|
Description: | Change point tests for joint distributions and copulas using pseudo-observations with multipliers or bootstrap. The processes used here have been defined in Bucher, Kojadinovic, Rohmer & Segers <doi:10.1016/j.jmva.2014.07.012> and Nasri & Remillard <doi:10.1016/j.jmva.2019.03.002>. |
Authors: | Bouchra R Nasri [aut], Bruno N Remillard [aut, cre, cph] |
Maintainer: | Bruno N Remillard <[email protected]> |
License: | GPL-3 |
Version: | 0.1.7 |
Built: | 2024-12-23 06:39:13 UTC |
Source: | CRAN |
Pseudo-observations used in Nasri, Remillard, Bahraoui (2021). The values represent conditional cdfs of Gaussian HMM models applied to log-returns of Nasdaq and Dow Jones Industrial indexes from 2007 and 2008. If the models are correct, the pseudo-observations should be almost iid with uniform distribution.
data(pseudos)
data(pseudos)
Pseudo-observations from Gaussian HMM models with 3 regimes for log-returns of the to Nasdaq index and Dow Jones Industrial indexes from 2007 and 2008.
1st column: pseudo-observations of a Gaussian HMM model with 3 regimes applied to the Nasdaq log-returns.
2nd column: pseudo-observations of a Gaussian HMM model with 3 regimes applied to the Dow Jones Industrial log-returns.
This function compute the Cramer-von Mises and Kolmogorov-Smirnov test statistics based on the new sequential process of Bucher et al (2014), using multipliers and parallel computing.
test.change.point( x, N = 1000, n_cores = 2, boot.method = "multipliers", est = FALSE )
test.change.point( x, N = 1000, n_cores = 2, boot.method = "multipliers", est = FALSE )
x |
(n x d) matrix of data (observations or pseudo-observations, including residuals), d>=1 |
N |
number of multipliers samples to compute the P-value |
n_cores |
number of cores for parallel computing (default = 2) |
boot.method |
bootstrapping method: 'multipliers' (default, fastest) or 'bootstrap' |
est |
if TRUE, tau is estimated (default = FALSE) |
CVM |
Cramer-von Mises statistic |
KS |
Kolmogorov-Smirnov statistic |
pvalueCVM |
Pvalue for the Cramer-von Mises statistic |
pvalueKS |
Pvalue for theKolmogorov-Smirnov statistic |
tauCVM |
Estimated changepoint using the Cramer-von Mises statistic |
tauKS |
Estimated changepoint using the Kolmogorov-Smirnov statistic |
Bouchra R Nasri and Bruno N Remillard, August 6, 2020
Nasri, B. R. Remillard, B., & Bahraoui, T. (2022). Change-point problems for multivariate time series using pseudo-observations, J. Multivariate Anal., 187, 104857.
x=matrix(rnorm(600),ncol=3) out = test.change.point(x)
x=matrix(rnorm(600),ncol=3) out = test.change.point(x)
This function compute the Cramer-von Mises and Kolmogorov-Smirnov test statistics based on the new sequential process of Bucher et al (2014), using multipliers and parallel computing. Two methods of bootstrapping are used: non-sequential (fastest) and sequential. Both methods yields basically the same P-valueas.
test.change.point.copula.BKRS( x, N = 1000, n_cores = 2, method = "nonseq", est = FALSE )
test.change.point.copula.BKRS( x, N = 1000, n_cores = 2, method = "nonseq", est = FALSE )
x |
(n x d) matrix of data (observations or pseudo-observations, including residuals), d >=2 |
N |
number of multipliers samples to compute the P-value |
n_cores |
number of cores for parallel computing (default = 2) |
method |
'nonseq' (default) or 'seq' |
est |
if TRUE, tau is estimated (default = FALSE) |
CVM |
Cramer-von Mises statistic |
KS |
Kolmogorov-Smirnov statistic |
pvalueCVM |
Pvalue for the Cramer-von Mises statistic |
pvalueKS |
Pvalue for theKolmogorov-Smirnov statistic |
tauCVM |
Estimated changepoint using the Cramer-von Mises statistic |
tauKS |
Estimated changepoint using the Kolmogorov-Smirnov statistic |
Bouchra R Nasri and Bruno N Remillard, August 6, 2020
Nasri, B. R. Remillard, B., & Bahraoui, T. (2022). Change-point problems for multivariate time series using pseudo-observations, J. Multivariate Anal., 187, 104857.
Bucher, A., Kojadinovic, I., Rohmer, T., & Segers, J. (2014). Detecting changes in cross-sectional dependence in multivariate time series, J. Multiv. Anal., 132, 111–128.
x<-matrix(rnorm(100),ncol=2) out = test.change.point.copula.BKRS(x)
x<-matrix(rnorm(100),ncol=2) out = test.change.point.copula.BKRS(x)