Re: Computing the covariance between regression coefficients

On 10 Jan 2006 20:52:21 -0800, "DanSoper@xxxxxxxxxxx"
<DanSoper@xxxxxxxxxxx> wrote:

>Hi Richard,
>
>Thank you for your prompt reply! The formula that you provided
>generates values that are very close to those contained in the
>asymptotic covariance matrices output by SPSS and SAS. I suspect that
>that the differences are due to rounding error, or perhaps SPSS and SAS
>use a bootstrapping approach to generating the ACOV matrix...
>
>If at all possible, could you provide me with a citation for the
>formula that you provided?
>
>Thanks again,
>
>-DAN

First, be sure you include Ray Koopman's correction. The formula I
gave applies only if the variance of the errors equals 1.0. Guess I
was sleepy.

I don't have a citation handy. I derived it from the matrix
expression.
-***
----------------------
Richard Startz RichardStartz@xxxxxxxxxxx
Lundberg Startz Associates
.