vineGoF: Vine Goodness-of-fit Tests

In vines: Multivariate Dependence Modeling with Vines

Description

Goodness-of-fit tests to verify whether the dependence structure of a sample is appropriately modeled by vine model.

Usage

vineGoF(vine, data, method = "PIT", ...)

Arguments

vine	A Vine object.
data	Data matrix of pseudo-observations.
method	Goodness-of-fit method. Supported values: "PIT" (Probability Integral Transform).
...	Additional arguments for the goodness-of-fit method.

Details

The "PIT" (Probability Integral Transform) method uses the vinePIT function to transform the data into variables which are independent and Uniform(0,1) and then use a hypothesis test to verify whether the resulting variables are independent and Uniform(0,1). The additional parameter statistic specifies the test to be applied for this purpose.

statistic

Statistic used to verify if the transformed variables are independent and Uniform(0,1). The default value is "Breymann" and supported methods are:

"Breymann"

Test proposed in the Section 7.1 of (Aas et al., 2009). See (Breymann et al., 2003) for more information.

Value

A vineGoF or a subclass with specific information about the goodness-of-fit method used. The statistic slot of this object contains the value of the statistic and pvalue the p-value.

References

Aas, K. and Czado, C. and Frigessi, A. and Bakken, H. (2009) Pair-copula constructions of multiple dependence. Insurance: Mathematics and Economics 44, 182–198.

Breymann, W. and Dias, A. and Embrechts, P. (2003) Dependence structures for multivariate high-frequency data in finance. Quantitative Finance 1, 1–14.

See Also

vineGoF, vinePIT.

Examples

copula <- normalCopula(c(-0.25, -0.21, 0.34, 0.51, -0.07, -0.18), 
                       dispstr = "un", dim = 4)
data <- rCopula(100, copula)

selectCopula <- function (vine, j, i, x, y) {
    data <- cbind(x, y)
    fit <- fitCopula(normalCopula(), data, method = "itau")
    fit@copula
}
normalCVine <- vineFit("CVine", data, method = "ml",
                       selectCopula = selectCopula,
                       optimMethod = "")@vine
normalDVine <- vineFit("DVine", data, method = "ml",
                       selectCopula = selectCopula,
                       optimMethod = "")@vine
show(normalCVine)
show(normalDVine)

normalCVineGof <- vineGoF(normalCVine, data, method = "PIT",
                          statistic = "Breymann")
normalDVineGof <- vineGoF(normalDVine, data, method = "PIT",
                          statistic = "Breymann")
show(normalCVineGof)
show(normalDVineGof)

vines documentation built on May 2, 2019, 5:55 a.m.