On 8 Jun 2007, at 12:09, Moss, Antony Charles wrote:
> ... though I
> know that there are some objections to the true 'randomness' of the
> random number function in Excel. There are forums online and
> everything- I found them by mistake, I swear...
:-)
There's a problem with computer-generated random numbers in general:
numbers spewed out by a program - i.e. a deterministic, executable
formula - aren't very random. They tend to be called pseudorandom
numbers. The important things is that they behave as if they were
truly random. What "truly random" means is not uncontroversial, but
most people tend not to lose much sleep over it.
R has a package <http://cran.r-project.org/src/contrib/Descriptions/
random.html> which uses random atmospheric noise supplied by <http://
www.random.org/> to generate the random numbers. There's lots of
other stuff on that web page for anyone interested in what random
really means. If you're really really interested then Knuth (1999,
pp 61-75) has some discussion of the properties you'd expect random
numbers to have, e.g. equidistribution (if you're using a uniform
generator), equidistributed pairs of numbers, you'd want the gaps
between numbers to have certain properties, there's something called
the poker test (how often do pairs of numbers occur, two pairs, three
of a kind, ...) - and so on...
Andy
Knuth, D. E. (1999). The Art of Computer Programming Volume 2:
Seminumerical Algorithms (3rd ed). Addison Wesley.
--
Andy Fugard, Postgraduate Research Student
Psychology (Room F15), The University of Edinburgh,
7 George Square, Edinburgh EH8 9JZ, UK
Mobile: +44 (0)78 123 87190 http://www.possibly.me.uk
|