Re: Basketballers observation was a signpost

Ray Murphy wrote:
  > RM: I find it strange that there is still no single formula in
> statistics for predicting random distributions, or any references to

If you can predict it, it isn't random. So, it should follow you can't 
have a formula that creates truly random distributions...You can only 
try for something that is "random enough".

> the G-test came closer to my tests with PC generated random data -
> although still not good enough.

Technically, computers don't/can't generate "random" data. (By their 
very nature they are an orderly bunch. <grin>) They can only generate 
pseudo-random data. If you generate enough of it for a long enough time, 
it repeats itself in a predictable pattern.

> One problem with trying to verify the value of the various tests that
> are used to show how far our data varies from normal distribution is
> the fact that some personal computer programs generate better random
> numbers than others.

True, the best random number generators are those used in cryptography, 
but even they are not truly random. The better the algorithm is at being 
"random" the harder it tends to be to program. Also, it tends to run 
slower the better it is as well.

Todd

-- 
Free Astrology Resources (FAR)
http://carnesoft.quotaless.com