The parameter of DCC model

[pascal]

I am estimating the DCC model with financial data. However, I obtained a negative alpha in the second stage of DCC estimation. Is it wrong? Does it violate the positive definiteness of DCC model? Is it right if I adjust the initial value of alpha and beta in order to fix this problem? If not, how can I solve it?
Many thanks
Pascal

[TomDoan]

There probably is no way to "fix" the problem other than not use DCC. If CC is adequate, then the correct value for the DCC alpha is zero (which means that in any given sample it can be slightly positive or slightly negative), and beta can (at least theoretically) be any value between zero and one.

[pascal]

Dear Tom,
Thank you for your reply.
Yes. The DCC alpha in my estimation is slightly negative. I will try to use other models.
I am appreciated with your help.
Many thanks

[billmil]

I just got RATS 10 and am estimating a DCC GARCH model, and I get an odd result:

garch(p=1, q=1, model=ar8, mv=dcc, rvectors=resid, vcv, hmatrices=hh, robust) / longap nygap sfgap singgap sydgap
## SX22. Expected Type SERIES[VECTOR[REAL]], Got SERIES[REAL] Instead


I have always used the rvectors=resid command before and had no problem-indeed I typed this in word-for-word (except the variable names) from a previous file where I had no problems. I would be most grateful for any help.
Thanks,
Bill

[TomDoan]

I don't recall any time when the target of RVECTOR could be anything other than a SERIES[VECT]. It looks like RESID is used someplace earlier in your program as a simple SERIES.