GJR(1,1)-MA(1)-M

[Alepruz]

Hello,

I was trying to replicate using RATS the example 9.8 of Taylor's book Asset Price Dynamics, Volatility, and Prediction.
On this one he does a GJR-GARCH(1,1) in mean with a MA(1) regressor on the mean.
The data he uses are returns of SP100 so I called the series RETURNSP.

If I would like to do only a GJR-GARCH(1,1) I would use the following code:

GARCH(P=1,Q=1,ASYMMETRIC, method=bhhh,REGRESSORS) / RETURNSP
# %GARCHV

Then I was thinking maybe I can add the MA(1) term in the mean the following way.

boxjenkins(DEFINE=M1,ma=1) RETURNSP

and then adding to the GARCH formula model=M1
such as:

GARCH(P=1,Q=1,ASYMMETRIC, method=bhhh,REGRESSORS,model=M1) / RETURNSP
# %GARCHV

But then the GJR doesn't work.

Do I need to set up the GARCH with lower level code instead of by using the GARCH function?

Thanks

Andres

[TomDoan]

Hello,

I was trying to replicate the example 9.8 of Taylor's book Asset Price Dynamics, Volatility, and Prediction in RATS.
On this one he does a GJR-GARCH(1,1) in mean with a MA(1) regressor on the mean.
The data he uses are returns of SP100 so I called the series RETURNSP.

If I would like to do only a GJR-GARCH(1,1) I would use the following code:

GARCH(P=1,Q=1,ASYMMETRIC, method=bhhh,REGRESSORS) / RETURNSP
# %GARCHV

Then I was thinking maybe I can add the MA(1) term in the mean the following way.

boxjenkins(DEFINE=M1,ma=1) RETURNSP

and then adding to the GARCH formula model=M1
such as:

GARCH(P=1,Q=1,ASYMMETRIC, method=bhhh,REGRESSORS,model=M1) / RETURNSP
# %GARCHV

But then the GJR doesn't work.

Do I need to set up the GARCH with lower level code instead of by using the GARCH function?

Thanks

Andres

What you're looking for is the %MVGAVGE series. What you probably want is:

Code: Select all

garch(p=1,q=1,asymmetric, method=bhhh,regressors) / returnsp
# constant %garchv %mvgavge{1}

Once you include the REGRESSORS option, the automatic CONSTANT is dropped from the model, so if you want the intercept in the mean equation, you need to include it yourself.