Stable Asymptotics for M-estimators

We review some first- and higher-order asymptotic techniques for M- estimators and we study their stability in the presence of data contaminations. We show that the estimating function (psi) and its derivative with respect to the parameter (grad psi) play a central role. We discuss in detail the first-order Gaussian density approximation, saddlepoint density approximation, saddlepoint test, tail area approximation via Lugannani-Rice formula, and empirical saddlepoint density approximation (a technique related to the empirical likelihood method). For all these asymptotics, we show that a bounded (in the Euclidean norm) psi and a bounded (e.g., in the Frobenius norm) grad psi yield stable inference in the presence of data contamination. We motivate and illustrate our findings by theoretical and numerical examples about the benchmark case of one-dimensional location model.