Volatility Forecast Toolbox (Scripts) Publisher's description
The Toolbox forecasts the volatility of a (mxn) vector of data and from a variety of in-built / non-in-built GARCH models with various distributions, as well as the univariate RiskMetrics.
The Toolbox forecasts the volatility of a (mxn) vector of data and from a variety of in-built / non-in-built GARCH models with various distributions, as well as the univariate RiskMetrics. The toolbox also estimates a number of Volatility Forecast Loss Functions and if selected Value-at-Risk for both short and long positions, along with a number of back-tests.
The three main functions are the following:
VFMatlab.m: In-built Matlab functions (i.e. models: GARCH, GJR-GARCH and EGARCH, distributions Normal and StudentвЂ™s t-Distribution)
VFGarch.m: Forecasts the volatility of the following models and distributions:
The supported models are:
GARCH, GJR-GARCH, EGARCH,
NARCH: Nonlinear ARCH: Engle and Bollerslev (1986)
NGARCH: Nonlinear GARCH: Higgins and Bera (1992)
AGARCH: Asymmetric GARCH: Engle (1990)
APGARCH: Asymmetric Power GARCH: Ding, Granger and Engle (1993)
NAGARCH: Nonlinear Asymmetric GARCH: Engle and Ng (1993)
The supported distributions are:
GAUSSIAN: Gaussian / Normal Distribution
T: Student t-Distribution
GED: Generalized Error Distribution
CAUCHY: Cauchy-Lorentz Distribution
HANSEN: Hansen's Skew-t Distribution
GC: Gram-Charlier Expansion Series
Finally, VFRiskMetrics.m forecasts the volatility of RiskMetrics model.
Furthermore, not only the rolling window (P) and maximum number of forecasts (max_forecast) can be specified but the number of forecasts that are of interest (int_forecasts). This means, for example, that one forecasts the volatility of daily returns for up to 22 days, but is interested only for the 1-day, 1- and 2-weeks and 1-month periods. Finally, a number of volatility forecast evaluation metrics are estimated and are given in: LossFunctions.m.
The forecast evaluation metrics are: Mean Square Error, Median Absolute Error, Median Absolute Percentage Error, Mean Absolute Deviation, Heteroskedasticity-adjusted Mean Square Error, Heteroskedasticity-adjusted Mean Absolute Error, R2LOG and QLIKE.
If it is enabled, Value-at-Risk for both long and short positions is estimated as well as a number of back-tests, which are given in VaRTest.m. The VaR are computed according to the distribution specified for the estimation of the GARCH model.
Value-at-Risk Backtesting for Long or/and Short Positions entail the estimation of the following: Percentage of Failures, Time Until First Failure, Christoffersen (2003) test for correct conditional coverage, which covers: unconditional coverage, independence test, and conditional coverage, and finally Basel II Accord evaluation criteria
System Requirements:MATLAB 7.6 (R2008a)
Program Release Status: New Release
Program Install Support: Install and Uninstall