The RATS Software Forum

Hi All,

I am trying to understand the relationship between oil prices (daily) and stock market returns (daily). I plan to use three different methods and from my reading i could get these can be the correct codes . Request for your confirmation about any changes needed and also how to make sure that this is the correct model specification in terms of lags and also correct diagnostic tests please. I am novice in this and hence pardon my ignorance (dummy variables are oil stock)

VAR (1) – BEKK – GARCH (1,1) Model

system(model=varmodel)
variables oil stock
lags 1
det constant
end(system)
garch(model=varmodel,mv=bekk,pmethod=simplex,piters=10,method=bfgs)

BEKK – GARCH (1,1) Model

garch(p=1,q=1,mv=bekk, pmethod=simplex, piters=10) / oil stock

DCC Garch

arch(p=1,q=1,mv=dcc) / oil stock

Are those in returns? (They should be).

Those are the correct codes for the models that you say you are estimating. Note that the first two nest---the VAR(1)-BEKK is just a BEKK model with a more complicated mean model. You use @MVQSTAT applied to the jointly standardized residuals to see if the model has taken care of the serial correlation. @MVARCHTEST applied to the jointly standardized residuals checks whether there is residual GARCH.

In practice, there is no "correct" GARCH model---they are all approximations. While BEKK and DCC don't nest, often one or the other will have a substantially worse log likelihood.