rugarch (version 1.0-3)

ugarchroll-methods: function: Univariate GARCH Rolling Density Forecast and Backtesting

Description

Method for creating rolling density forecast from ARMA-GARCH models with option for refitting every n periods and some multicore parallel functionality.

Usage

ugarchroll(spec, data, n.ahead = 1, forecast.length = 500, refit.every = 25, 
refit.window = c("recursive", "moving"), parallel = FALSE, 
parallel.control = list(pkg = c("multicore", "snowfall"), cores = 2), solver = "solnp", 
fit.control = list(), solver.control = list(), calculate.VaR = TRUE, 
VaR.alpha = c(0.01, 0.05), ...)

Arguments

spec

A univariate GARCH spec object specifiying the desired model for testing.

data

A univariate dataset.

n.ahead

The number of periods to forecast.

forecast.length

The length of the total forecast for which out of sample data from the dataset will be excluded for testing.

refit.every

Determines every how many periods the model is re-estimated.

refit.window

Whether the refit is done on an expanding window including all the previous data or a moving window, the length of the window determined by the argument above (refit.every).

parallel

Whether to make use of parallel processing on multicore systems.

parallel.control

The parallel control options including the type of package for performing the parallel calculations (multicore for non-windows O/S and snowfall for all O/S), and the number of cores to make use of.

solver

The solver to use.

fit.control

Control parameters parameters passed to the fitting function.

solver.control

Control parameters passed to the solver.

calculate.VaR

Whether to calculate forecast Value at Risk during the estimation.

VaR.alpha

The Value at Risk tail level to calculate.

...

.

Value

• An object of class uGARCHroll.

Details

GARCH models generate a partially time varying density based on the variation in the conditional sigma and mean values (skewness and shape are usually not time varying in GARCH models unless the underlying distribution has an interaction with the conditional sigma). The function first generates rolling forecasts of the ARMA-GARCH model and then rescales the density from a standardized (0, 1, skew, shape) to the one representing the underlying return process (mu, sigma, skew, shape). Given this information it is then a simple matter to generate any measure of risk through the analytical evaluation of some type of function of the time varying density. The function calculates one such measure (VaR), but since the full time varying density parameters are returned, the user can calculate many others (see for example partial moments based measures or the Pedersen-Satchell family of measures). The argument refit.every determines every how many periods the fit is recalculated and the total forecast length actually calculated. For example, for a forecast length of 500 and refit.every of 25, this is 20 windows of 25 periods each for a total actual forecast length of 500. However, for a refit.every of 30, we take the floor of the division of 500 by 30 which is 16 windows of 30 periods each for a total actual forecast length of 480 (16 x 30). The important thing to remember about the refit.every is that it acts like the n.roll argument in the ugarchforecast function as it determines the number of rolls to perform. For example for n.ahead of 1 and refit.every of 25, the forecast is rolled every day using the filtered (actual) data of the previous period while for n.ahead of 1 and refit.every of 1 we will get 1 n.ahead forecasts for every day after which the model is refitted and reforecast for a total of 500 refits (when length.forecast is 500)! The function has 2 main methods for viewing the data, a standard plot method and a new report methods (see class uGARCHroll for details on how to use these methods).

See Also

For specification ugarchspec, fitting ugarchfit, filtering ugarchfilter, forecasting ugarchforecast, simulation ugarchsim, parameter distribution and uncertainty ugarchdistribution, bootstrap forecast ugarchboot.

Examples

data(sp500ret)
ctrl = list(rho = 1, delta = 1e-9, outer.iter = 100, tol = 1e-7)
spec = ugarchspec(variance.model = list(model = "sGARCH", garchOrder = c(1,1)), 
		mean.model = list(armaOrder = c(0,0), include.mean = TRUE), 
		distribution.model = "std")

sp500.bktest = ugarchroll(spec, data = sp500ret, n.ahead = 1, 
forecast.length = 100, refit.every = 25, refit.window = "recursive", 
solver = "solnp", fit.control = list(), solver.control = ctrl, 
calculate.VaR = TRUE, VaR.alpha = c(0.01, 0.025, 0.05))
report(sp500.bktest, type="VaR", n.ahead = 1, VaR.alpha = 0.01, 
conf.level = 0.95) 
report(sp500.bktest, type="fpm")