Jackknifing fixed points of iterations

Abstract

The jackknife method for estimating the dispersion of a statistic T , considered as an estimator of a p -dimensional vector of parameters, can be modified so as to become computationally feasible for the class of statistics that are obtained as limits of iterative processes. Examples are given to show how the method can be applied to the EM algorithm and to iteratively reweighted leastsquares.