Some Stopping times for Stochastic Approximation Procedures.
FLORIDA STATE UNIV TALLAHASSEE DEPT OF STATISTICS
Pagination or Media Count:
Sequential bounded length confidence interval procedures are developed for stochastic approximation procedures of the Robbins-Monro type. Sufficient conditions for the confidence intervals to have asymptotically the prescribed confidence coefficients and the stopping times to be asymptotically efficient are given. The behavior of some of these procedures in several Monte Carlo experiments is briefly described. Author
- Statistics and Probability