Q1. Section of 12.1.5 of RATS User’s Guide suggest the ARCH-X and GARCH-X models.
“Add regressors for the mean model and for the variance model”
When we have these additional variables, do we need to check for the Multi-collinearity? That is, do we need to check whether the additional regressors are correlated just like the OLS case?
Multi-collinearity in the the GARCH-X models?
-
garchrookie
- Posts: 11
- Joined: Sat Jan 03, 2009 11:37 pm
Re: Multi-collinearity in the the GARCH-X models?
How you choose variables for the mean model is going to depend quite a bit on what you're trying to do.
Is your main interest the GARCH part of the model? In that case, the point of the regressors is to correct for identifiable mean shifts which would (if not extracted) cause false readings on the size of the residuals. Your regressors would likely be very carefully selected variables (things like dummy variables for the day of the week) that are unlikely to have collinearity problems.
Is your main interest the regression part of the model? If so, you'll have to choose the regressors that you need for that model. If they're highly correlated, you may have to live with that unless you can find a transformation which will clean that up a bit.
Is your main interest the GARCH part of the model? In that case, the point of the regressors is to correct for identifiable mean shifts which would (if not extracted) cause false readings on the size of the residuals. Your regressors would likely be very carefully selected variables (things like dummy variables for the day of the week) that are unlikely to have collinearity problems.
Is your main interest the regression part of the model? If so, you'll have to choose the regressors that you need for that model. If they're highly correlated, you may have to live with that unless you can find a transformation which will clean that up a bit.