# How to determine the optimal lag order of asymmetric innovation for a GARCH?

In the python package arch_model, there is an option with which we can give the "lag order of the asymmetric innovation" when we estimate GARCH model. Are there any ways to find the optimal lag order besides looking at a number of models with different lag orders for the asymmetric innovation via AIC or BIC?

• What type of model do you want to estimate ? When looking at a standard GARCH model, I don't know what "asymmetric innovation" means. Do you want to estimate an asymmetric GARCH model like TGARCH, GJR-GARCH etc. ? – Lars May 15 at 7:44
• I would want to estimate GJR-GARCH, but I am not sure whether "asymmetric innovation" here would be equivalent to that. I can not find any other information about this parameter. – Aqqqq May 15 at 10:17
• In this context asymmetric innovation means how many lags of the term $I \cdot \epsilon_{t-i}$ you want to include in the variance equation. $I$ is an indicator variable which takes the value 1 if $\epsilon_{t-i}<0$ and zero otherwise. – Lars May 15 at 12:32
• Just asking: hat is the source of your information? Sounds exactly like GJRGARCH for me. But still how should I choose the optimal lag? – Aqqqq May 15 at 16:08
• *what is the source of your information? – Aqqqq May 15 at 19:20