Inequalities and extremal problems in probability and statistics. Selected topics (Q2833249)

This volume presents six articles on important inequalities in probability and statistics, and related extremal problems. These include uniform and non-uniform Berry-Esseen bounds and inequalities for martingales, random polynomials and generalized cross-moments. Each of these articles presents the state of the art in the relevant area, together with some new techniques, results or applications.NEWLINENEWLINEThe first of these articles, written by Osȩkowski, presents applications of the method of moments to proving inequalities for martingales using a dynamic programming approach and backwards induction. Various kinds of inequalities are examined in this general framework, including bounds for the maximal value of a martingale difference sequence, bounds for the sum of squares of a martingale difference sequence, a prophet inequality for \(L^2\)-bounded martingales, and an inequality for martingale transforms. NEWLINENEWLINEThe second article is also by Osȩkowski. Here, the author studies optimal stopping problems for a Brownian motion \((B_t)_{t\geq0}\) and its maximal function \((B^*_t)_{t\geq0}\), where \(B^*_t=\sup_{0\leq s\leq t}|B_s|\). In particular, the author considers the non-integral setting, deriving upper bounds on the weak Lorentz norms \(\lVert B^*_\tau\rVert_{L^{q,\infty}}\), where \(q\geq1\), \(\tau\) is a stopping time, andNEWLINE\[CARRIAGE_RETURNNEWLINE\lVert\xi\rVert_{L^{q,\infty}}=\sup_{\lambda>0}\left[\lambda^q\mathbb{P}(|\xi|\geq\lambda)\right]^{1/q}\,.CARRIAGE_RETURNNEWLINE\] NEWLINEThese upper bounds are given in terms of weak and strong norms of \(B_\tau\). NEWLINENEWLINEAn article by Shevtsova begins with a thorough overview of the best known constants in the Berry-Esseen inequality for sums of independent and IID random variables, and in several variants of non-uniform Berry-Esseen inequalities. Some new non-uniform inequalities are then derived.NEWLINENEWLINEPinelis then presents two new methods for obtaining non-uniform Berry-Esseen inequalities, using a novel combination of Fourier and extremal techniques and smoothing inequalities. Each of these methods is illustrated with a proof of a classical non-uniform Berry-Esseen bound. NEWLINENEWLINEIn the fifth article, Pinelis presents new uniform Berry-Esseen bounds for the self-normalized sumNEWLINE\[CARRIAGE_RETURNNEWLINE\frac{X_1+\cdots+X_n}{\sqrt{X_1^2+\cdots+X_n^2}}\,,CARRIAGE_RETURNNEWLINE\]NEWLINEwhere \(X_1,\ldots,X_n\) are independent random variables with mean zero. The upper bounds here are expressed in terms of third, fourth and sixth moments of the \(X_i\), since allowing the use of higher moments than the third results in improvements in the constants appearing in the upper bounds.NEWLINENEWLINEFinally, de la Peña and Ibragimov derive inequalities for random polynomials (including sample autocovariances and sample cross-moments as special cases). Their method also applies in the case of dependence between the underlying random variables. Applications in various statistical testing settings are discussed.