Comparison of estimators of the fraction defective in the normal distribution

Abstract

Two point estimators of the survival fraction are compared using Pitman's measure of closeness. Expressions for obtaining the conditional mean absolute errors are derived. The “odds” that one estimator is closer to the true value forms the basis for determining the preferred estimator in the normal failure model which is often used in industrial quality control.

Keywords:

• Pitman closeness
• conditional mean adsolute error
• maximum likelihood and minimum variance unbiased estimators
• suevival probabilities