Hi All,
The recent discussions about how often the basketballers scores could
occur by chance motivated me to make up a little program to find out
exactly what was happening by testing with random numbers..
The Chi square theory in statistics says it was a 30 in 1000 chance
and
Hermes was saying a graph like that could ocur 100 in 1000 times by
chance but my experiments have shown repeatedly that (contrary to
established statistical theory) it is actually about 2.25 per 1000.
The experiment involved generating random numbers from 0.000 to
0.9999 and placing a score of 1 in the first graph line for all
numbers
under 0.5000 and placing a score of 1 in the second graph line for all
numbers over 0.5000 in sets of 502 random numbers, representing
502 basketballers.
4.000 tests were done and there were 9 cases out of 4,000 where the
score was as good or better than the basketballers score I got (with a
gap of 68 or more between the low and high score) which is 2.25 per
1000.
A second and third seies of 4000 tests showed another 9 and 7cases.
Summary:
* Hermes ..............100 per 1000
* Statistical theory... 30 per 1000 -- (Chi square and p-score)
* Experiment(1) ...2.25 per 1000 = 9/4000
* Experiment(2)....2.25 per 1000 = 9/4000
* Experiment(2)....1.75 per 1000 = 7/4000
This information needs to be independently verified with separate
programming of random data but it would appear that the Chi
square formula is quite useless in the range for several hundred
factors of 2 types - such as the 502 basketballers.
Astrologers who are using Chi square could easily be throwing
away interesting results if they are not aware of this.
Ray
