[R] Testing for randomness

[Prof. Brian Ripley]

Autocorrelations are by definition linear.  Do you mean non-linear 
dependencies in a bivariate scatterplot?

There are many such tests, but R is not the plase to find them.  Pierre 
l'Ecuyer has a test suite, and George Marsaglia's DIEHARD can still be 
found.


On Fri, 28 Mar 2003 pfm401 at lineone.net wrote:

> Dear all,
> 
> Is there a test in R for the randomness of a sequence of observations (e.g.
> to test the random number generator)? Specifically I am looking for autocorrelations
> which are not necessarily linear in nature, which the acf function does
> not seem to be flexible enough to detect as it tests for linear autocorrelation.
> 
> Thanks in advance,
> Paul.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595