bashtage / arch

ARCH models in Python
Other
1.29k stars 247 forks source link

How to modify GARCH model to incorporate new terms ? #709

Closed a412h closed 6 months ago

a412h commented 6 months ago

I have read the documentation, but I cannot find a way to implement the following modification to the GJR GARCH model: $\sigmat^2 = \omega + \beta \sigma{t-1}^2 + \gamma \epsilon{t-1}^2 I{ [ \epsilon{t-1} < 0] } + \eta \epsilon{t-1}^2I{ [ \epsilon{t-1} > 0] } .$ How do I implement the model above ?

bashtage commented 6 months ago

The model you have is observationally equivalent to a standard GJR_GARCH.

$\sigmat^2 = \omega + \beta \sigma{t-1}^2 + \gamma1 \epsilon{t-1}^2 I{ [ \epsilon{t-1} < 0] } + \eta \epsilon{t-1}^2I{ [ \epsilon_{t-1} > 0] } .$

is the same as

$\sigmat^2 = \omega + \beta \sigma{t-1}^2 + \gamma2 \epsilon{t-1}^2 I{ [ \epsilon{t-1} < 0] } + \alpha \epsilon_{t-1}^2 .$

where $\eta=\alpha$ and $\gamma_2 + \alpha = \gamma_1$.