Estimation of non-normal GARCH model
The function ugarchfit()
does a joint estimation of all the mean, variance and distribution parameters. A general approach is to use a skewed student t distribution. You then need to estimate also the skew
and shape
parameters \(\xi\) and \(\nu\).
In this exercise, you fit the GARCH model with skewed student t distribution on a simulated return series called ret
. The true model used to simulate has the following parameters
list(mu = 0, ar1 = 0, ma1 = 0, omega = 6*10^(-7), alpha1 = 0.07, beta1 = 0.9,
skew = 0.9, shape = 5)
You will see that you obtain parameter estimates are close to the true parameters. The difference between the estimated and true parameter is called the estimation error. On long time series, the error is typically small.
This exercise is part of the course
GARCH Models in R
Exercise instructions
- Plot the return series
ret
and note the large negative return. - Complete the instructions to specify a GARCH model with skewed student t distribution.
- Estimate the model.
- Extract the coefficients from the obtained
ugarchfit
object.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Plot the return series
___
# Specify the garch model to be used
garchspec <- ___(mean.model = list(armaOrder = c(0,0)),
variance.model = list(model = "sGARCH"),
___ = ___)
# Estimate the model
garchfit <- ___(data = ___, spec = ___)
# Inspect the coefficients
___(___)