Absolute deviations of the binomial about a prescribed centre: the tail expansion in closed form

RESUMO

Let $X\sim Bin(N,p)$ and let $c=Nr$ be a prescribed centre with $r\ne p$. The finite formula for $E|X-c|$ is classical, and the existence of a complete large-deviation expansion in powers of $N^{-1}$ follows from Timashev's saddle-point expansion for binomial tails. We give the coefficients in closed form for the stop-loss tail produced by the absolute deviation, with the lattice defect of $Nr$ retained explicitly. The construction combines three standard ingredients: Bernoulli polynomials for the logarithm of successive binomial masses, the exponential Bell recursion, and Eulerian polynomials for weighted geometric sums. The $k$-th coefficient has denominator exactly $(1-\rho)^{2k+2}$, where $\rho$ is the exponential tilt; hence the effective expansion parameter is $1/(N(1-\rho)^2)$, the lattice weighted analogue of the parameter in the incomplete gamma expansion. The case $r=\tfrac12$ gives the fold at the origin of a biased $\pm1$ random walk, and its parity split is identified with the lattice defect of the fold.