[R] A function for random test based on longest run (UNCLASSIFIED)

[Dirk Eddelbuettel]

Brian,  

On 29 December 2007 at 17:03, Prof Brian Ripley wrote:
| On Thu, 27 Dec 2007, Dirk Eddelbuettel wrote:
| 
| >
| > On 27 December 2007 at 12:08, bogdan romocea wrote:
| > |   > require(tseries)
| > |   > ?runs.test
| > | Also, take a look at dieharder, it implements a large number of
| > | randomness tests:
| > | http://www.phy.duke.edu/~rgb/General/dieharder.php
| >
| > Also note that CRAN has an RDieHarder package that provides access to
| > DieHarder tests from R.  RDieHarder is currently happier on Linux than on
| > Windows. If anybody wants to contribute build instructions for Windows ...
| 
| It's not easy on Linux!  There is a src/Makefile that ignores most of the 
| info from how R was built, including CC, CFLAGS andlibrary pats such as 
| /usr/local/lib64.  

Ack, that needs fixing for amd64 etc. The rest worked for us...

| And that's if you can get dieharder built: it gave me 
| lots of errors before succeeding.  

FWIW, Robert, DieHarder's principal author (CC'ed), works on a Fedora system,
and we know of successful ports to (at least) BSD.  So unless you tell us how
it fails there is little we can do.

| I managed to build from the SRPM, but 
| that installed with the include files without read permissions.
|
| But I've succeeded on Windows,

Excellent!  I had one quick look a few month ago on my MinGW setup at work
and ... didn't have enough (and interest) to make it work.  But having a
working RDieHarder on Windows would be very useful, so thanks!

| and RDieHarder is now in the CRANextras 
| collection (and so can be installed from the menus etc). I'll send Dirk 
| separately notes on what I had to do to make it work.

I look forward to receiving those.  We also got patches from Anton
Korobeynikov (CC'ed) for building DieHarder on Windoze but I haven't had time
to look at those.

Thanks, Dirk

-- 
Three out of two people have difficulties with fractions.