From Steve Hunt-Anschutz
I suspected that I could test the validity of Benford’s law by examining a
large set of random numbers each chosen from a random range. I reasoned that to
choose a representative random range I needed to select from a log base 10 of
ranges. This would avoid skewing towards larger collections of digits.
Thereafter, a number could be chosen from each range linearly at random. So I
sat at my UNIX terminal and typed the following single line command:
yes | sed `100000q’ | awk `BEGIN{srand();u=6*log(10)}
{printf”%e\n”,rand()*exp(rand()*u)}’ | cut -c1 | sort | uniq-c
This tries 100 000 numbers. The lower range limit is always 0, the upper
between 1 and 1 000 000. And it worked! Now I know why the smaller digits always
go first from the house-numbering section of my local DIY store.
sha@prosumsw.com
