baq_nifunction: News Impact on Conditional Variance

In sebinum/baqgarchutil: Analyse a Fitted baqGARCH Model

Description

Apply a news impact Function to a mGJR class object.

Usage

baq_nifunction(x, epsnames = c("series1", "series2"), er_grid = 10,
  er_grid_by = 0.2, ni_long = TRUE, ni_wide = TRUE, quiet = FALSE)

Arguments

x	A mGJR class object.
epsnames	Custom names for the variables eps1 and eps2. Character vector of length two. Defaults to c("eps1", "eps2").
er_grid	The minimum and maximum value for the hypothetical returns in the news impact Function. Numerical vector of length one.
er_grid_by	Steps by which to construct the sequence of returns. Numerical vector of length one.
ni_long	Logical switch if the conditional variance after news impact should be returned in long format (as data.frame).
ni_wide	Logical switch if the conditional variance after news impact should be returned in wide format (as matrix).
quiet	Logical switch if information about applying baq_nifunction should be shown in the console.

Value

The applied news impact Function packaged as a baq_nif class object. The news impact functions' conditional variance / correlation (optional) as long- (data.frame) and and wide-format (list of matrices). The values are defined as:

series_names

character vector with the names for eps1 and eps2

eps

data.frame with eps1 and eps2 inherited from the mGJR class object.

baq_h

data.frame with the estimated conditional covariance matrices inherited from the mGJR class object. Each column stands for a matrix element of the conditional covariance matrix, each row for the time at which it was observed. The time notation of the rows is identical to eps.

coef

data.frame with the baqGARCH coefficients needed for the news impact Function.

coef_se

data.frame with the baqGARCH coefficients standard errors.

coef_tval

data.frame with the baqGARCH coefficients T-values.

baq_ni_cndh_long

data.frame containing the news impact on the conditional variance / correlation of eps1 / eps2 in long-format.

baq_ni_cndh_wide

A list of matrices with the news impact on the conditional variance of eps1/eps2 and conditional correlation in wide-format.

Details

Schmidbauer & Roesch (2008, 2014) proposed a bivariate asymmetric quadratic GARCH (baqGARCH) model to account for asymmetric components in bivariate volatility matrices. They also apply a news impact on the conditional volatility of a fitted baqGARCH model by letting the conditional volatility matrices H = (h_{ij}) depend on x = (x_1, x_2):

x -> H(x) = C'C + A'xx'A + B'∑B + S_w(x) * Γ'xx'Γ

where ∑ is the unconditional covariance matrices of the bivariate time series and x the vector of potential innovations (i.e. returns) in the bivariate series affecting it's conditional volatility.

The contour lines (conditional variance after news impact) are based on the functions:

x -> h_{11}(x), x -> h_{22}(x), x -> h_{12} (x)/√(h_{11}(x) * h_{22}(x)),

where the function x_{11} stands for the news impact on the next day's conditional variance of returns on series 1, x_{22} stands for the news impact on the next day's conditional variance of returns on series 2 and h_{12}(x)/√{h_{11}(x) * h_{22}(x)} for the conditional correlation of returns (series 1 & 2).

References

Schmidbauer, H. & Roesch, A. (2008). Volatility Spillovers Between Crude Oil Prices. International Conference on Policy Modeling. EcoMod, Berlin.

Schmidbauer, H. & Roesch, A. (2014). Volatility Spillovers Between Crude Oil Prices and Us Dollar To Euro Exchange Rates. 4th IAEE Asian Conference,Beijing.

See Also

baq_niplot to plot the news impact function and diag_mv_ch_model to perform tests to determine model adequacy

Examples

# create data
eps <- mgarchBEKK::simulateBEKK(2, 100)

# fit the model
gjr <- mgarchBEKK::mGJR(eps$eps[[1]], eps$eps[[2]])

# apply the news impact function to the model
nif <- baq_nifunction(gjr)

sebinum/baqgarchutil documentation built on May 8, 2019, 11:58 p.m.