The RATS Software Forum

Dear Tom,

when calculating the conditional correlations form the DCC models I understood that we have to use the garch command and save the hmatrices=hh and then use set rho12 = hh(t)(1,2)/sqrt(hh(t)(1,1)*hh(t)(2,2))
and that this is the same for all mvgarch models. however for the dcc model aren't we supposed to use the fitted values (the q) from the correlation equation as rho.

regards
Hashem

hashem wrote:Dear Tom,

when calculating the conditional correlations form the DCC models I understood that we have to use the garch command and save the hmatrices=hh and then use set rho12 = hh(t)(1,2)/sqrt(hh(t)(1,1)*hh(t)(2,2))
and that this is the same for all mvgarch models. however for the dcc model aren't we supposed to use the fitted values (the q) from the correlation equation as rho.

regards
Hashem

They're the same thing. The HH(T)(1,2) is computed using the combination of the rho's and the HH(T)(1,1) and HH(T)(2,2). The formula used in the SET applies to all types of models, not just DCC which is why we use it.

Thank you Tom for your prompt reply as always.

Can I ask your advice about another issue. when i run the DCC model in most cases (of countries i use) the DCC(1) is always insignificant does that mean i cant rely on the conditional correlations to reach conclusions about the linkk between markets? I understand that if both are insignificant then the model collapses to a constnat correltion one.

regards
Hashem