Re: testing equality of covariances
- From: Richard Ulrich <Rich.Ulrich@xxxxxxxxxxx>
- Date: Tue, 20 Sep 2005 17:40:02 -0400
On Mon, 19 Sep 2005 19:49:21 EDT, Jack Tomsky <jtomsky@xxxxxxxxxxxxx>
wrote:
> > I would like to test the equality of covariances (not
> > covariance matrices) for two independent samples. The
> > only tests I have found are for covariance matrices.
> > Can anyone steer me in the right direction (offer
> > references, etc)? Many thanks
> >
> > Jerry
>
> Are these two bivariate populations? That is, does each population have a single covariance, Cov(X,Y)?
>
> In that case, if you can assume that Var(X) and Var(Y) are the same for both populations, then it reduces to testing for the equality of two correlation coefficients.
I answered the question yesterday, where it was also posted in
sci.stat.consult.
In short - I thought it was a badly expressed hypothesis,
and I suggested the test of regression coefficients in
the combined sample. The hypothesis seemed silly if
it wasn't the same two variables in different samples --
though I did not consider two variables correlated with
one, as Ray suggested.
The test on regressions is usually preferable to the test
of two correlations, since the latter has to assume
the equal variances, as Jack says.
>
> If the sample sizes are sufficiently large, you can use Fisher's z-statistic.
- that's referring to the test on correlations.
--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.
- References:
- Re: testing equality of covariances
- From: Jack Tomsky
- Re: testing equality of covariances
- Prev by Date: K-L distance
- Next by Date: Computing Really Tiny p-values using Fisher's Exact Test
- Previous by thread: Re: testing equality of covariances
- Next by thread: Re: testing equality of covariances
- Index(es):
Relevant Pages
|
Loading
