From: Paul Gilmartin Sent: Friday, 2 November 2012 3:37 PM
I'm missing something here; most likely an unstated assumption. My intuition agrees with test program making a histogram of consecutive integers mod 3:
SPPG@MVS3:139$ cat foomod /* Rexx */ signal on novalue; /* Doc: Demonstrate hash clustering. */ Counts. = 000 /* Set all counters to 0. */
do I = 1 to 99999 /* Generate some samples. */ R = I // 36 /* mod 36 */ Counts.R = Counts.R + 1; end
do I = 0 to 35 /* Display counters. */ say 'Counts.'right(I, 2, 0 ) '=' right( Counts.I, 4 ); end
Which produces output:
SPPG@MVS3:140$ foomod Counts.00 = 2777
...
Counts.35 = 2777
Here I see no evidence of clustering.
There won't be. All the data is uniformly distributed. Real data is not uniform.