Abstract:
A process is described whereby the Fabian modification of the Robbins — Monro algorithm successively approaches an isolated stable root of the regression function. A conditional limit distribution is determined of normalized deflections of the successive approximations from this root provided that the process converges to it.