Truncated sequential tests for the circular error probability

Abstract

Suppose shots are fired sequentially and R is the radius of the circle (centred at the expected point of impact of a given shot with probabiltiy . Then R is known as the circular error probability. A general class of truncared at m sequential procedures based on asymptotic confidence sequences for R is constructed for testing R≦R₀ (acceptable) against (not acceptable) where is a given standard. The error probabilities are asymptotically bounded by given outside an indifference region with . Asymptotic properties and asymptotic efficiencies are considered for special cases. The performance of some procedures for normal and contaminated normal distribution is studied by Monte Carlo Methods.

Keywords:

• confidence sequences
• robust estimation of dispersion
• regularly varying function
• asymptotic relative efficiency
• Monte Carlo Methods