Asymptotic theory of sequential estimation: Differential geometrical approach (Q1175393)

The authors analyze the higher order asymptotic properties of sequential estimation procedures in multiparameter curved exponential families. They use the differential geometrical approach to show that suitably chosen stopping times improve various asymptotic characteristics of the estimators. Some of the stopping times in question are those ones for which the so-called exponential curvature is reduced to 0. Therefore, the present paper clarifies also why finding stopping times for which the initial curved exponential families are reduced to noncurved exponential ones is important, in the asymptotic sense. For a discussion of the latter problem in exponential families of stochastic processes the reader is referred to the reviewer's paper, Statistical inference from stochastic processes, Proc. AMS-IMS-SIAM Jt. Summer Res. Conf., Ithaca/NY 1987, Contemp. Math. 80, 323-330 (1988; Zbl 0672.62084).