Mean squared prediction errors
The GJR GARCH model is a generalization of the GARCH model. It should thus lead to a better fit in terms of lower Mean Squared Errors (MSE). Let's verify this on the Microsoft returns msftret for which garchfit corresponds to the estimation with the standard GARCH(1,1) model, while gjrfit is when the GJR model is used. Remember that you can compute the vector with prediction errors \(e\) for the mean using the residuals() method. The prediction error for the variance equals the difference between \(e^2\) and the predicted GARCH variance.
This exercise is part of the course
GARCH Models in R
Exercise instructions
- Compute the vector with prediction errors for the means using the
residuals()method. - Complete the code for calculating the MSE for
garchfitestimation output. - Compute the MSE for the
gjrfitestimation output.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Compute prediction errors
garcherrors <- ___(garchfit)
gjrerrors <- ___(gjrfit)
# Compute MSE for variance prediction of garchfit model
___((___(garchfit)___ - garcherrors^2)___)
# Compute MSE for variance prediction of gjrfit model
___