Abstract:It is well known that fluctuations of marginals of high-dimensional random vectors that satisfy a certain concentration estimate called the thin shell condition are approximately Gaussian. In this article we identify a general condition on a sequence of high-dimensional random vectors under which one can identify the exponential decay rate of large deviation probabilities of the corresponding sequence of marginals. More precisely, consider the projection of an n-dimensional random vector onto a random k n -dimensional basis, k n ≤ n , drawn uniformly from the Haar measure on the Stiefel manifold of orthonormal k n -frames in R n , in three different asymptotic regimes as n → ∞ : “constant” ( k n = k ), “sublinear” ( k n → ∞ but k n / n → 0 ) and “linear” ( k n / n → λ with 0 < λ ≤ 1 ). When the sequence of random vectors satisfies a certain “asymptotic thin shell condition”, we establish large deviation principles for the corresponding sequence of random projections in the constant regime, and for the sequence of empirical measures of the coordinates of the random projections in the sublinear and linear regimes. We also establish large deviation principles for scaled ℓ q norms of the random projections in all three regimes. Moreover, we show that the asymptotic thin shell condition holds for various sequences of random vectors of interest, including the uniform measure on suitably scaled ℓ p n balls, for p ∈ [ 1 , ∞ ) , and generalized Orlicz balls defined via a superquadratic function, as well as a class of Gibbs measures with superquadratic interaction potential. Along the way, we obtain logarithmic asymptotics of volumes of high-dimensional Orlicz balls, which may be of independent interest. We also show that the decay rate of large deviation probabilities of Euclidean norms of multi-dimensional projections of ℓ p n balls, when p ∈ [ 1 , 2 ) , exhibits an unexpected phase transition in the sublinear regime, thus disproving an earlier conjecture due to Alonso-Gutiérrez et al. Random projections of high-dimensional random vectors are of interest in a range of fields including asymptotic convex geometry and high-dimensional statistics.

Estimation of the directional parameter of the offset exponential and normal distributions in three-dimensional space using the sample mean

Deformed normal distributions

A Stein Goodness-of-fit Test for Directional Distributions

Directional Statistics and Filtering Using libDirectional

Bounds for distributional approximation in the multivariate delta method by Stein's method

Parameter Estimation and Model-Based Clustering with Spherical Normal Distribution on the Unit Hypersphere

Estimation of nominal direction of arrival and angular spread using the determinant of the data matrix

Two-Dimensional Angular Parameter Estimation for Noncircular Incoherently Distributed Sources Based on an L-Shaped Array

Recent advances in directional statistics

The normal law under linear restrictions: simulation and estimation via minimax tilting

Applications of Sequentially Estimating the Mean in a Normal Distribution Having Equal Mean and Variance

Generalized T 3 -plot for testing high-dimensional normality

Stein's Method of Moments on the Sphere

Estimating the mean and variance of a high-dimensional normal distribution using a mixture prior

Trigonometric Moments of a Generalized von Mises Distribution in 2-D Range-Only Tracking

Deviation bounds for the norm of a random vector under exponential moment conditions with applications

Nonparametric Measure-Transportation-Based Methods for Directional Data

An asymptotic thin shell condition and large deviations for random multidimensional projections

Skewed Bernstein-von Mises theorem and skew-modal approximations

An asymptotic expansion of the empirical angular measure for bivariate extremal dependence

Generalized inverse transformation method via representative points in statistical simulation