From: "Weiqi Gao" <weiqigao AT crl DOT com>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: random #'s
Date: 2 Dec 1996 13:23:05 GMT
Organization: CRL Dialup Internet Access
Lines: 30
Message-ID: <01bbe053$bf5d70e0$010200c0@weiqigao>
References: <01BBDE40 DOT 57785E00 AT sky38 DOT dataplusnet DOT com> <329FF108 DOT 7ED4 AT cs DOT com> <32a28934 DOT 29578785 AT news DOT ox DOT ac DOT uk>
NNTP-Posting-Host: crl10.crl.com
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp

George Foot <mert0407 AT sable DOT ox DOT ac DOT uk> wrote in article
<32a28934 DOT 29578785 AT news DOT ox DOT ac DOT uk>...
> On Sat, 30 Nov 1996 00:32:08 -0800, "John M. Aldrich"
> <fighteer AT cs DOT com> wrote:
> 
> >Michael Matczynski wrote:
> >> 
> >> I wrote this program and it proves that it is an accurate random
number generator.
> >
> >Gee, thanks for crediting me for the code that I gave you.
> 
> Heh... It's always good to see credit given where it's due.
> 
> Anyway, it's not really very useful... To really test the randomness,
> you should perform more sophisticated tests, such as the Chi-squared
> test, the Kolmogorov-Smirnov test, a less crude frequency test, and
> also some form of serial test, gap test, "poker" test,
> "Coupon-collector's test" (I love these names...), permutation test,
> run test, ... there are many.
> 
> Which of these you bother with would depend on what you will be using
> the results for, of course...

Of cause, for random number generators, tests are not enough.  A
mathematical proof is a must.

-- 
Weiqi Gao
weiqigao AT crl DOT com