Binomial coefficients with non-integer arguments

When n and r are positive integers, with n ≥ r, there is an intuitive interpretation of the binomial coefficient C(n, r), namely the number of ways to select r things from a set of n things. For this reason C(n, r) is usually pronounced “n choose r.”

But what might something like C(4.3, 2)? The number of ways to choose two giraffes out of a set of 4.3 giraffes?! There is no combinatorial interpretation for binomial coefficients like these, though they regularly come up in applications.

It is possible to define binomial coefficients when n and r are real or even complex numbers. These more general binomial coefficients are in this liminal zone of topics that come up regularly, but not so regularly that they’re widely known. I wrote an article about this a decade ago, and I’ve had numerous occasions to link to it ever since.

The previous post implicitly includes an application of general binomial coefficients. The post alludes to coefficients that come up in Bessel’s interpolation formula but doesn’t explicitly say what they are. These coefficients B_k can be defined in terms of the Gaussian interpolation coefficients, which are in turn defined by binomial coefficients with non-integer arguments.

$\begin{eqnarray*} G_{2n} &=& {p + n - 1 \choose 2n} \\ G_{2n+1} &=& {p + n \choose 2n + 1} \\ B_{2n} &=& \frac{1}{2}G_{2n} \\ B_{2n+1} &=& G_{2n+1} - \frac{1}{2} G_{2n} \end{eqnarray*}$

Note that 0 < p < 1.

The coefficients in Everett’s interpolation formula can also be expressed simply in terms of the Gauss coefficients.

$\begin{eqnarray*} E_{2n} &=& G_{2n} - G_{2n+1} \\ F_{2n} &=& G_{2n+1} \\ \end{eqnarray*}$

Leave a Reply