Functional calculus and asymptotic theory for statistical analysis (Q1812785)

Let \({\mathcal F}\) be a convex class of distribution functions on the real line containing all degenerate distributions, and let \(T\) be a functional defined on \({\mathcal F}\). Further, let \(\{ X_ 1,\dots,X_ n\}\) be a sample from a distribution function \(F\) in \({\mathcal F}\), with empirical distribution function \(F_ n\). The author studies properties of estimators \(T(F_ n)\) of \(T(F)\). Special attention is paid to exponential-rate bounds for the probability \(P(| T(F_ n) - T(F)| > t)\), Bahadur representations, convergence rate of \(T(F_ n) - T(F)\) to the normal distribution and asymptotic representations for bootstrap type statistics. The proofs are based on a functional calculus approach assuming smoothness properties on the functional \(T\), such as locally Lipschitz continuity and locally Lipschitz differentiability with respect to some norm on \({\mathcal F}\). The second part of the paper concerns \(M\)- and \(L\)-estimators. These estimators can be viewed as special cases, i.e., for some specific functional \(T\) they can be written as \(T(F_ n)\). Sufficient conditions are derived under which the previously derived results hold for these types of statistics. Specifically, sufficient conditions for locally Lipschitz differentiability of the functional are given.