At 20/9/2001 18:00 -0700, Rodger Whitlock wrote:
>Actually, most people, when presented with truly random data, see
>patterns in it. You can demonstrate this very easily in a spreadsheet
>program. There should be a simple function that gives uniformly
>random numbers in the interval 0-1. Set up two columns for X and Y
>coordinates of points in a scatter chart, then fill them up with the
>random number function (in Lotus it used to be simply "@random").
>Look at the chart that results. You will "see" clumps and clusters
<start of rant>
Since the numbers are generated by programs they are NOT random. In the
early days of computers plots as described above of these "pseudo random"
numbers showed diagonal white lines. Doubtless the programs producing them
are nowadays more complex, but the programs are still programs and the
numbers are still NOT random.
However even truly random scattering leads to "clumping". This was notable
in an early method of colour photography in which coloured specs were
scattered on the unexposed photographic plate. (This was the work of the
Truly random numbers are used in the British "premium bond" lottery, but
the technique counts decays of carbon 14 atoms.
In the years leading up to WW2 such numbers were produced by manually
observing numbers around the edge of a wheel through a slit that only
showed one digit. The wheel was illuminated by an erratically flashing neon
light and written down by hand. I think I still have a book with a table of
so-constructed random numbers.
The draw for British "premium bonds" is made by counting decays of carbon
14 atoms. The numbers so produced are theoretically random and have passed
all tests of randomness to which they have been exposed.
I have reason to believe that the "one time pads" used by the British
government for high security messages were calculated by the above techniques.
Further details will be found in back numbers of the Journal of the Royal
Statistical Society. I can't remember whether they appeared in Series A
(general) or Series B (theoretical).
<end of rant>
I'm not in touch with recent statistical work, but the early computer
generated "random numbers" were not considered suitable for many practical
David Ibbetson * 133 Wilton Street * Unit 506 * Toronto M5A 4A4
mailto:[log in to unmask]
Phone:(416)363-6692 Cel:(416)831-6692 Fax:(416)363-4987