Training covariance down

[Guest guest]

Hi Pete,

I ws just reading an old note of yours about covariance. When there is

a lot of covariance in one bar, I know that one option is simply to

train that signal down. I also have a design from you called cov_dvd.

Does this design train the standard deviation? Do you (or anyone else

out there who has done this) have a preference?

In terms of a decision tree (training tone issues first, etc.), where

do you place covariance?

Thanks,

Jane