The RATS Software Forum

Hi,

I have been studing many different statistic packages and so far I didn't find any which could solve my problem.
That's why I decided to write on this forum.
I am trying to estimate univariate GARCH(4,0) model with 4 exogenous variables as below:
Mean equation: X=0
Conditional equation: h= constant+ a1*hetero1(-1)+a2*hetero1(-2)+a3*hetero1(-3)+a4*hetero1(-4)+b1*h(-4)

So in conditional variance equation I have 4 exogenous variables (hetero1 with different lags) and 1 previous value of conditional variance but lagged by 4 days.

The problem is that any software can perform optimization.

Do Rats can solve that problem?
I was browsing automatic window for ARCH/GARCH and I couldn't find the restrictions option. If there be such option, then I will estimate GARCH(4,0) with exogenous variables and restricted GARCH(-1), GARCH(-2) and GARCH(-3) parameters.

Could you please help me with that?

Thanks in advance,
Joanna

You would have to use MAXIMIZE to handle that. I'm a bit confused about your model, though. You seem to have a "GARCH" model without any "ARCH", that is, there is no lagged squared residual term. Is that correct?

Hi Tom,

Yes, it looks like I am not using squared residuals. But I am following P. Molnar "High-low range in GARCH models of stock return volatility, where I am introducing Parkinson estimators to conditional variance equations as exogenous variables instead.

Is there any Rats programs for GARCH with exogenous variables and restrictions on parametrers (GARCH(-1 to -3)), or RATS allows to estimete just 4th lag of GARCH proces, without estimating lag first, second and third?

Thanks for your help!
Joanna

Nothing that specific. However, the Tsay textbook has quite a few non-standard GARCH models with skipped lags and/or exogenous shifts. tsayp155.rpf and tsayp158.rpf would be the closest to what you're doing.