Bayesian Tolerance Intervals for Zero-Inflated Data with Applications in Pharmaceutical Quality Control

摘要

Quality control is an essential operation of the pharmaceutical industry. Tolerance intervals are commonly used in quality control to assure that a large proportion of critical quality attributes meet certain specification criteria with high confidence. Discrete measurements; such as microbial cell count; number of defective devices or discretized continuous responses due to limit of quantitation or rounding; are abundant in pharmaceutical manufacturing and quality control. Tolerance intervals for discrete variables are widely used in industrial applications to set alert and action limits for critical quality attributes for the purpose of control and surveillance. A substantial amount of discrete data consist of excess number of zeros and often displays over-dispersion in variance. By far; there is no statistical method available.