Re: [Diehard] Overlap sum test

From: Ernst Lippe (ernstl-at-planet-dot-nl_at_ignore.this)
Date: 09/26/03 

Date: Fri, 26 Sep 2003 20:59:00 +0200

On Wed, 24 Sep 2003 08:59:43 +0000, Cristiano wrote:

> Ernst Lippe wrote:

>> So far so good. Each individual y is the sum of a set of
>> i.i.d. variables, so from the central limit theorem its distribution
>> must approach the normal distribution when no_sum is sufficiently
>> large.
>>
>> [interesting]
>>
>> I have not done the actual calculations to determine if this
>> non-normality would have a real influence on the outcomes of the test,
>> but I still find it somewhat strange. I would expect that a
>> non-linear transformation would give better results.
>
> Why don't you have done the actual calculations? I done many tests to write
> what I wrote and they have taken much time.
> I'd like to hear some comment based on real facts. The theory is beautiful,
> but the practice is wonderful! :-)

But I gave you a fact, if I understand Marsaglia's post correctly
I pointed you to the source of the problem.

>> On the other hand, I don't expect that this is the source of your
>> problems. Marsaglia is an expert so it seems reasonable to assume that
>> his approximation is OK.
>
> Well, in this case don't touch a thing.
> I totally disagree when I hear this kind of argument. Nobody is infallible.

OK, you were certainly right in this case.

BTW, thanks for your efforts, the Diehard battery of tests are highly
useful, and it is important to have a good implementation.

greetings,

Ernst Lippe

Relevant Pages

  • Re: [Diehard] Overlap sum test
    ... so from the central limit theorem its distribution ... >> I have not done the actual calculations to determine if this ... > I'd like to hear some comment based on real facts. ...
    (sci.crypt)
  • Re: Blind Cable Test at CES
    ... Well, I am a statistician. ... Before looking at your calculations, we need to consider something that's ... binomial distribution in that case. ... sample is large enough to use a normal approximation in the first place.) ...
    (rec.audio.opinion)
  • Re: Blind Cable Test at CES
    ... Well, I am a statistician. ... Before looking at your calculations, we need to consider something that's ... binomial distribution in that case. ... sample is large enough to use a normal approximation in the first place.) ...
    (rec.audio.opinion)
  • Re: Is Itos Lemma correct?
    ... your calculations? ... What algorithm did you use for your random number generator? ... Here's the original location and distribution of Mr. Steve Park: ... It's characterized as "a very accurate approximation of the normal idf". ...
    (sci.math)
  • Re: Crosscorrelation problems
    ... cross-correlation calculations. ... correlation would then be 1. ... the absence of temporal autocorrelation, the correlation coefficient, ... distribution has mean zero and a variance of ^-1. ...
    (sci.stat.consult)