A combinatorial identity
Posted by tpc at March 23rd, 2009
I came across the following identity today.

It is not exactly difficult to establish if one uses the Pascal relation repeatedly.


An alternative proof is to extract the coefficient of
from the generating function identity:

I’m thinking that this must have a nice combinatorial proof, but all my references are in my office.

One combinatorial interpretation is as follows: if you are picking (k 1) out of m items, given some ordering of the m items you can break your choices up into disjoint cases based on how many of the items you pass up before choosing the first. After choosing the jth item (and passing up the first j-1) you have exactly (m - j) \choose k ways to choose the rest.
Brent Yorgey
Thanks Brent. That was neat.
tpc