Robust Garch(1,1) Model

A method for modeling robust generalized autoregressive conditional heteroskedasticity (Garch) (1,1) processes, providing robustness toward additive outliers instead of innovation outliers. This work is based on the methodology described by Muler and Yohai (2008) .

robustGarch

robustGarch is an R package aiming to provide a method for modelling robust Garch processes (RG), addressing the issue of robustness toward additive outliers - instead of innovations outliers. This work is based on Muler and Yohai (2008) (MY).

Installation

The package can be installed as following:

devtools::install_github("EchoRLiu/robustGarch")
library(robustGarch)

Example

This is a basic example which shows you how to fit your daily return time series data into robust Garch(1,1) model.

if (requireNamespace("PCRA", quietly = TRUE)) {
  library(robustGarch)
  
  ret <- PCRA::retOFG
  ret <- ret$RET
  
  (robFitBM <- robGarch(ret, fitMethod = "BM"))
  
  sum(robFitBM$fitted_pars[2:3])
  summary(robFitBM)
  plot(robFitBM)
} else {
  message("PCRA package is not installed. Please install it with install.packages('PCRA') if you want to run this example or use other dataset to replace ret.")
}

For more examples and explanation, please refer to the robustGarch-Vignette.

Future Development

Any future development will be released in the github page. A few key features will be added to the package in September 2020:

• Fix the issue with singularity error with Hessian matrix
• Statistics tests such as std_error, t_value, p_value for Garch parameters
• Code debug on model filter for M model and QML
• More optimization choices
• Extension to robust Garch(p, q)
• Name changes for better collaboration