The RATS Software Forum

Hi
I am estimating ADCC GARCH correlations using the attached code. I have estimated correlation for all pairs in my dataset. For a few pairs, I am getting constant correlation values for every week (I am using weekly returns data).

I have attached the code, data file and output of one pair as a specimen (see xls file for correlation results). Could you please guide me why are correlation values same for every week and where I am going wrong.

I appreciate your help.

Thank you!

Best regards

Your second step ADCC estimates don't look very good. (In effect, it's estimating a CC model, hence the behavior). The residuals have a fairly low correlation to start, but I would look really carefully at the individual diagnostics for the univariate models, as it looks like an ADCC model isn't appropriate.

TomDoan wrote:You second step ADCC estimates don't look very good. (In effect, it's estimating a CC model, hence the behavior). The residuals have a fairly low correlation to start, but I would look really carefully at the individual diagnostics for the univariate models, as it looks like an ADCC model isn't appropriate.

Thanks Tom!

I am now attempting to identify the best fit model (DCC, ADCC, GDCC or AGDCC) on the basis of BIC criterion for all pairs in my time series data set. Hopefully this step will help.

I would do some careful checks on the individual GARCH models. That's almost looking like the results that you would get from having one enormous outlier (or typo) in one of the series.