DEN Discussion List Archive
[Date Prev][Date Next][Date Index]
[Thread Index]
[Author Index]
Re: Random Numbers
- Subject: Re: Random Numbers
- From: John <jsdwd@ispwest.com>
- Date: Mon, 18 Oct 2004 21:07:58 +0700
- Replyto: jsdwd@ispwest.com
- User-agent: Microsoft-Entourage/11.0.0.040405
Most systems for generating random numbers generate pseudo-random numbers.
There is an algorithm used to generate the number based on a seed. Where
the seed comes from and what algorithm is used determines the degree to
which thenumbers produced can be considered random. There was a wide range
of results. I haven't kept up and don't know the current state of these
things. Perhaps a competent statistician and not an old hack like me could
comment on the present state of these things.
People who produce random number generators are, for the most part, aware of
the problems inherent in this undertaking and use different techniques to
make their number more (ahem) random. Some don't bother. So part of the
answer to your (seemingly) simple question is that it depends what you want
to use the numbers for. How random do you need them to be? (Is that like
asking if one can be a little bit pregnant?)
Since we cannot know if a given system produces random numbers (some of them
haven't been produced yet), we need to operationalize. This is usually done
using a criterion set forth in a test.
There are such tests for randomness in a series of numbers. I know that Dr.
Lloyd S. Nelson published a fair amount on this subject for his Technical
Aids for the journal known as the Journal of Quality Technology. ASQC
published a compilation (or two) of his technical aids. It may still be
available. His late son Peter also published on the subject I believe.
John Dowd
jsdwd@ispwest.com
DEN Home |
Main Index |
Thread Index |
Author Index