A Berry–Esséen Bound for M‐estimators (Q4376537)

Let \(X\) be a random variable taking values in a measurable space \(({\mathfrak R},{\mathfrak I})\) with distribution \(F\). For an open set \(T\) of \(R^1\), let \(f:{\mathfrak R}\times T\to R^1\) be a jointly measurable function. Suppose that \(X_1, \dots,X_n\) are independent observations drawn from the distribution \(F\). A random variable \(t_n=t_n(X_1, \dots,X_n)\) which minimizes the function \(n^{-1}\sum^n_1 f(X_i,t)\) is called an \(M\)-estimator. The authors prove a Berry-Esséen bound for general \(M\)-estimators under optimal regularity conditions on the score function \(f(x,t)\) and the underlying distribution \(F\). As an application, Berry-Esséen bounds are obtained for the sample median, the \(L_p\)-median, \(p>1\), and Huber's estimator of location.