Two kinds of iterative estimators, based on Robbins‐Monro stochastic approximation algorithms, are investigated. It has been shown that, under a set of sufficient conditions, the estimators are robust over a known class of distributions in asymptotically min‐max sense.For illustration, we have examined the robust estimators for the ?‐normal family of distributions. Numerical and simulation results show that, for both large and small sample sizes, the robust estimators are more efficient than the sample mean estimator for those nearly Gaussian distributions.
關聯:
Journal of the Chinese Institute of Engineers=中國工程學刊 2(2), pp.105-113
