Abstract:We consider estimating the shared mean of a sequence of heavy-tailed random variables taking values in a Banach space. In particular, we revisit and extend a simple truncation-based mean estimator by Catoni and Giulini. While existing truncation-based approaches require a bound on the raw (non-central) second moment of observations, our results hold under a bound on either the central or non-central $p$th moment for some $p > 1$. In particular, our results hold for distributions with infinite variance. The main contributions of the paper follow from exploiting connections between truncation-based mean estimation and the concentration of martingales in 2-smooth Banach spaces. We prove two types of time-uniform bounds on the distance between the estimator and unknown mean: line-crossing inequalities, which can be optimized for a fixed sample size $n$, and non-asymptotic law of the iterated logarithm type inequalities, which match the tightness of line-crossing inequalities at all points in time up to a doubly logarithmic factor in $n$. Our results do not depend on the dimension of the Banach space, hold under martingale dependence, and all constants in the inequalities are known and small.

Moment and exponential estimation for the distribution of the norms for random matrices martingales

Norm estimates in Grand Lebesgue Spaces for some operators, including magic square matrices

On the Almost Sure Central Limit Theorem for Vector Martingales: Convergence of Moments and Statistical Applications

Lindeberg's Central Limit Theorems for Martingale-Like Sequences under Sub-Linear Expectations

Exponential Inequalities for Martingales with Applications

Rates in the central limit theorem for random projections of Martingales

The Moment of Maximum Normed Randomly Weighted Sums of Martingale Differences

Limit Theorems and Inequalities Via Martingale Methods

Tail estimations for normed sums of centered exchangeable random variables

Large deviation inequalities for martingales in Banach spaces

Exponential inequalities for martingales with applications The following exponential inequality for supermartingales can be found in Freedman

Mean Estimation in Banach Spaces Under Infinite Variance and Martingale Dependence

Bernstein Type Exponential Inequalities for Martingales with Applications

Non-Uniform Berry-Esseen Bounds for Martingales with Applications to Statistical Estimation

Large Deviation Exponential Inequalities for Supermartingales

Bernstein-type and Bennett-type inequalities for unbounded matrix martingales

Optimal Estimates for Martingale Transforms

Factorizable convergence of random variables in Grand Lebesgue Spaces

Concentration and moment inequalities for heavy-tailed random matrices

Mixed Lp estimates for transforms of noncommutative martingales

Matrix concentration inequalities via the method of exchangeable pairs