Re: random numbers and statistics

From: Scott Wilber (swilber_at_comscire.com)
Date: 08/08/04 

Date: 8 Aug 2004 10:19:19 -0700

sean_incali@yahoo.com (sean kim) wrote in message news:<4d47f9d3.0408061611.7a4436c@posting.google.com>...
> Hello group,
>
> I'm doing soem simulations using random numbers, and i wanted to do
> statistics on the results of nuemrical calculations that were done
> using random number inputs.
>
> It's basically perturbation analysis with random and finite
> perturbations to the input parameters. then i'm looking at the
> fluctuations of values of a given output due the perturbations.
>
> Naturally, since the numbers are not really random, but pseudorandom,
> I'm wondering if the pseudo random generator will show false
> correlations if i look at thing like the mean, variance and standard
> deviations of the output.
>
> Am I right in worrying about this? or anyone has some insights they
> are graciously willing to provide?
>
>
> Thanks all in advance
>
> sean

The Mersenne Twister is one of the best prngs commonly available. We
have tested it continuously with our RNGmeter (free download at
http://www.comscire.com/Downloads/Files/RNGmeter/RNGmeterSetup.exe )
to about 2Tbytes and found only a slight bias in one of the
autocorrelation orders. This is much better than most other prngs that
fail catastrophically at a much lower byte count. If you use the
Twister, be sure that you get the version with the corrected seeding
algorithm. For example:
http://www-personal.engin.umich.edu/~wagnerr/MersenneTwister.html

Scott Wilber
Don't Leave Randomness to Chance.
ComScire.

Relevant Pages

  • Re: random numbers and statistics
    ... (random number generators) ... > perturbations to the input parameters. ... > fluctuations of values of a given output due the perturbations. ... > I'm wondering if the pseudo random generator will show false ...
    (sci.math.num-analysis)
  • Re: random numbers and statistics
    ... > statistics on the results of nuemrical calculations that were done ... > perturbations to the input parameters. ... > fluctuations of values of a given output due the perturbations. ... > I'm wondering if the pseudo random generator will show false ...
    (sci.math)
  • random numbers and statistics
    ... I'm doing soem simulations using random numbers, ... perturbations to the input parameters. ... fluctuations of values of a given output due the perturbations. ... I'm wondering if the pseudo random generator will show false ...
    (sci.math.num-analysis)
  • random numbers and statistics
    ... statistics on the results of nuemrical calculations that were done ... fluctuations of values of a given output due the perturbations. ... I'm wondering if the pseudo random generator will show false ...
    (sci.math)