The RATS Software Forum

Hi,
Has anyone implemented the DECO model of Engle and Kelly(2012), "Dynamic Equicorrelation", JBES, vol 30, no 2, 212-228.

Thank you

This uses one of the standard RATS example data files. (Not a great choice for the technique, since the Canadian exchange rate has a low correlation with the others).

Hi,

Is it possible to form the minimum variance and global minimum variance portfolios as done in the DECO paper (http://papers.ssrn.com/sol3/papers.cfm? ... id=1354525) using matrix algebra in RATS?

Thanks!

Hi all,


Does anybody know how to even output the covaraince matrix? I assume its the same as for the DCC model is anyone knows that?

Thanks!

xray wrote:Hi all,

Does anybody know how to even output the covaraince matrix? I assume its the same as for the DCC model is anyone knows that?

Thanks!

The version of this just posted now computes the covariance matrices as the SERIES[SYMMETRIC] called HH.

xray wrote:Hi,

Is it possible to form the minimum variance and global minimum variance portfolios as done in the DECO paper (http://papers.ssrn.com/sol3/papers.cfm? ... id=1354525) using matrix algebra in RATS?

Thanks!

Their process would take a lot of programming. Calculating the portfolios is simple given the covariance matrices. However, in order to avoid biasing the results, they run rolling estimates, but in order to avoid the cost of running too many optimization problems, they re-estimate only once in 22 periods, and do a sequence of 22 one-step predictions, using the last set of estimates. They also cut out the description of how to forecast the "rho" out-of-sample from an earlier working paper. It isn't analytical and they gave two different expressions, so it's not clear which they recommend.

TomDoan wrote:

The version of this just posted now computes the covariance matrices as the SERIES[SYMMETRIC] called HH.

Thanks.

Is it possible to display this in the matrix form

HH =
hh11............hh1,8
...................
...................
hh1,8...........hh8,8

The print and display commands don't seem to work

There is one HH matrix per time period. You can do

DISPLAY HH(100)

to get the HH matrix at T=100, and the same for any other data point.