Re: Statistics of cross-correlation

On Sun, 25 May 2008 16:07:21 -0700 (PDT), nico
<romano.nicola@xxxxxxxxx> wrote:

If you wish to use R you probably should consider moving this question
to the R-help mailing list. See the link Mailing lists a the R home
page.

I'm not a statistician, and I never used cross-correlation before. I
wrote into this list because I would like to understand the
statistical basis for analysis of cross-correlation. I don't care
about software specific issues (at least at the moment....).

"Significance" does not mean "importance"; it merely
means "unlikely to happen by chance". In fact, that can
be hard to avoid for Ns that get large enough.

I know what significance mean, but, I repeat, the test in the page
that was linked in the first post gives me a significant (so unlikely
to happen by chance) result no matter what values I test it for... I
can cross-correlate two random time series and still get a significant
cross-correlation.
So, if that is merely a problem of big Ns, is there a way to
circumvent it?

Always significant?

That won't happen for two "random time series" if it
is a real test and it is being applied appropriately.

Now, if your series both show a linear increase (decrease),
you will get some positive correlation, no matter the lag.
Is that what you are referring to?



- and if you are looking at one particular correlation,
or one lag, the others have *nothing* to do with it, until
you explicitly compute a correlation that partials-out
the other lags/ steps.

I'm sorry, I don't get this. What do you mean with "a correlation that
partials-out the other lags"?
I guess my confusion is coming from the fact that the Box-Ljung test
considers the other lags...

Like I said, I don't "do" time series. I don't know
the Box-Ljung test, or other special things for time series.
A general trend invalidates an assumption of "stationary
series," which may be the root of your trouble.


The problem is that I have not managed to find a simple and clear
explanation of this.
I would be grateful to anyone who could help me understand this
problem.
[...]

--
Rich Ulrich

http://www.pitt.edu/~wpilib/index.html
.