acdfilter-methods: ACD Model Estimation

In racd: Autoregressive Conditional Density Models

Description

Filtering of data with ACD dynamics.

Usage

acdfilter(spec, data, out.sample = 0,  n.old = NULL, skew0 = NULL, 
shape0 = NULL, ...)

Arguments

data	A univariate xts data object (or one which can be coerced to such).
spec	A univariate ACD spec object of class ACDspec.
out.sample	A positive integer indicating the number of periods before the last to keep for out of sample forecasting.
n.old	For comparison with ACDfit models using the out.sample argument, this is the length of the original dataset.
skew0	Optional recursion starting parameter for the skew dynamics. If not used, the transformed skew dynamics intercept value is used.
shape0	Optional recursion starting parameter for the shape dynamics. If not used, the transformed shape dynamics intercept value is used.
...	.

Value

A ACDfilter object containing details of the ACD filter.

Author(s)

Alexios Ghalanos

Examples

## Not run: 
# Check that fit/filter return the same values
library(racd)
library(rugarch)
data(sp500ret)
spec = acdspec(variance.model=list(variance.targeting = TRUE),
mean.model=list(armaOrder=c(1,1)),distribution.model=list(model = "jsu",
skewOrder=c(1,1,0), shapeOrder=c(1,1,0)))
fit = acdfit(spec, sp500ret)
# remove variance targeting:
spec = acdspec(variance.model=list(variance.targeting = FALSE),
mean.model=list(armaOrder=c(1,1)),distribution.model=list(model = "jsu",
skewOrder=c(1,1,0), shapeOrder=c(1,1,0)))
setfixed(spec)<-as.list(coef(fit))
# fit@fit$skhEst contains the untransformed recursion 
# starting values for the skew and shape dynamics.
# fit@model$sbounds contain the: [skew.LB, skew.UB, shape.LB, 
# shape.UB, shape.rate(for exponential transformation)]
# skew0 and shape0 take the transformed values (so use the 
# functions logtransform and exptransform)
filt = acdfilter(spec, sp500ret, skew0 = logtransform(fit@fit$skhEst[1], 
fit@model$sbounds[1],fit@model$sbounds[2]), 
shape0 = exptransform(fit@fit$skhEst[2], fit@model$sbounds[3], 
fit@model$sbounds[4], rate = fit@model$sbounds[5]))
head(cbind(kurtosis(fit), kurtosis(filt)))
head(cbind(skewness(fit), skewness(filt)))
head(cbind(sigma(fit), sigma(filt)))
head(cbind(quantile(fit, probs=c(0.01)), quantile(filt, probs=c(0.01))))

## End(Not run)

racd documentation built on May 2, 2019, 4:47 p.m.