18
1

Complete Asymptotic Expansions and the High-Dimensional Bingham Distributions

Abstract

For d2d \ge 2, let XX be a random vector having a Bingham distribution on Sd1\mathcal{S}^{d-1}, the unit sphere centered at the origin in Rd\R^d, and let Σ\Sigma denote the symmetric matrix parameter of the distribution. Let Ψ(Σ)\Psi(\Sigma) be the normalizing constant of the distribution and let Ψd(Σ)\nabla \Psi_d(\Sigma) be the matrix of first-order partial derivatives of Ψ(Σ)\Psi(\Sigma) with respect to the entries of Σ\Sigma. We derive complete asymptotic expansions for Ψ(Σ)\Psi(\Sigma) and Ψd(Σ)\nabla \Psi_d(\Sigma), as dd \to \infty; these expansions are obtained subject to the growth condition that Σ\|\Sigma\|, the Frobenius norm of Σ\Sigma, satisfies Σγ0dr/2\|\Sigma\| \le \gamma_0 d^{r/2} for all dd, where γ0>0\gamma_0 > 0 and r[0,1)r \in [0,1). Consequently, we obtain for the covariance matrix of XX an asymptotic expansion up to terms of arbitrary degree in Σ\Sigma. Using a range of values of dd that have appeared in a variety of applications of high-dimensional spherical data analysis we tabulate the bounds on the remainder terms in the expansions of Ψ(Σ)\Psi(\Sigma) and Ψd(Σ)\nabla \Psi_d(\Sigma) and we demonstrate the rapid convergence of the bounds to zero as rr decreases.

View on arXiv
Comments on this paper