One approach would be to do a permutation test: Look
at the difference in correlations that you are concerned
about compared to the distribution of differences of
correlations when the order of the data are randomly
permuted.
There are a couple of functions on the Burns Statistics website
for permutation tests that are associated with the working paper
"Permuting Super Bowl Theory" (about the outcome of the Super
Bowl predicting the stock market). However, I think that you
would need to adapt one of them to this application, or probably
more easily just start from scratch with a little "for" loop.
Patrick Burns
Burns Statistics
patrick@burns-stat.com
+44 (0)20 8525 0696
http://www.burns-stat.com
(home of S Poetry and "A Guide for the Unwilling S User")
Stahel, Christof wrote:
Hi all,
I know this is not a specific S+ question, but I thought someone knows
of an easy solution. If I were to test pairwise calculated correlations
over two subsamples, how do I test whether they are equal.
To be more precise, let x = cbind(x1,x2) be a matrix of dim (T,2). I
would like to test if cor(x[1:n,1],x[1:n,2]) =
cor(x[n+1:T,1],x[n+1:T,2]).
Any suggestions are very welcome.
Thanks
Christof
--------------------------------------------------------------------
This message was distributed by s-news@lists.biostat.wustl.edu. To
unsubscribe send e-mail to s-news-request@lists.biostat.wustl.edu with
the BODY of the message: unsubscribe s-news
|