Subject: Birthday paradox (was Re: RelCache (aka ELF prebinding) news)
To: Andrew Brown <atatat@atatdot.net>
From: Dave Huang <khym@azeotrope.org>
List: tech-kern
Date: 12/03/2002 13:16:56
On Tue, Dec 03, 2002 at 12:07:13PM -0500, Andrew Brown wrote:
> certainly there are characteristics that affect when people are born
> (yes, there are a lot less people with a birthday of february 29), but
> do you have any hard data that suggests the spread of birthdays over
> days of the year is sharply slanted in any direction?

I don't know if this is considered "sharply slanted," and it's a
self-selected sample, but here's a list of almost 1000 birthdays, along
with a histogram of their distribution down at the bottom of the page:
http://www.vulpine.pp.se/furbirth/

The most popular month is October, with about 10.5% of the birthdays,
and the least popular is January, with about 6.7%.

> the math is simple, if repetitive.  try it.

There's an approximation which is less repetitive to calculate:
  1 - e^((-k)*(k-1)/2n)
where n is the total number of possibilities (365 for the standard
birthday problem) and k is the number of people. The approximation
gets better for larger values of n. You can also solve for the 50%
point, and get:
  k = .5 + sqrt(1 + 8n ln(2))/2

which is often approximated even further to
  k = sqrt(2 ln(2)) * sqrt(n)

or about 1.18 * sqrt(n)
-- 
Name: Dave Huang         |  Mammal, mammal / their names are called /
INet: khym@azeotrope.org |  they raise a paw / the bat, the cat /
FurryMUCK: Dahan         |  dolphin and dog / koala bear and hog -- TMBG
Dahan: Hani G Y+C 27 Y++ L+++ W- C++ T++ A+ E+ S++ V++ F- Q+++ P+ B+ PA+ PL++