Hello,

Does anyone know why mgarch dcc predicted correlations begin with unreasonable values? For example, the following code results in correlation that begin at about 0 and slowly make their way to "reasonable" levels.

Code:
. use http://www.stata-press.com/data/r15/stocks
(Data from Yahoo! Finance)

. quietly mgarch dcc (toyota nissan honda = , noconstant), arch(1) garch(1)

. predict H*, correlation
Is there a way to force Stata to begin with something more reasonable, for example sample average?

Thank you,
Stan