auto_garch: Fit the Best GARCH Model to an Univariate Time Series

In Reckziegel/PortfolioMoments: Functions to be used in conjuction with PortfolioAnalytics

Description

This function searches over different model specifications to find the best according to one of the selection criterias: Akaike, Bayes, shibata, Hannan-Quinn and likelihood.

Usage

auto_garch(R, variance = c("sGARCH", "eGARCH", "gjrGARCH", "apARCH",
  "csGARCH"), distributions = c("norm", "std", "ged", "snorm", "sstd",
  "sged", "jsu", "ghyp"), garch_p = c(0, 1), garch_q = c(0, 1),
  arma_p = c(0, 1), arma_q = c(0, 1), criteria = c("Akaike", "Bayes",
  "Shibata", "Hannan-Quinn", "likelihood"), n.ahead = 1,
  conditional = TRUE, ...)

Arguments

R	A vector, matrix, data.frame, xts, timeSeries, zoo or a tibble object.
variance	A vector or a list of character strings with the variance models to be computed. It can be any combination of: "sGARCH", "eGARCH", "gjrGARCH", "apARCH", "csGARCH".
distributions	A vector or a list of character strings with the variance models to be computed. It can be any combination of: "norm", "std", "ged", "snorm", "sstd", "sged", "jsu", "ghyp".
garch_p, garch_q	A vector or list with the number of mimimum and maximum number of lags to be included in the garch process.
arma_p, arma_q	A vector or list with the mimimum and maximum number of lags to be included in the arma process.
criteria	The criteria in which the models will be evaluated. One of: "Akaike", "Bayes", "Shibata", "Hannan-Quinn" and "likelihood".
n.ahead	The number of periods ahead from which the sigmas should be forecasted.
conditional	TRUE or FALSE. If TRUE, the the conditional sigmas covariances is returned. If FALSE, the unconditional covariance is printed.
...	Any other parameters to pass thought ugarchspec.

Details

This function searchs thought the best GARCH models by "brute force". Since fitting a Garch's can be demanding, it is advisable to not explore all the options for portfolio optimization unless you have time and a enough memory into your computer.

Value

A variance-covariance matrix.

Reckziegel/PortfolioMoments documentation built on May 29, 2019, 1:21 p.m.