Confidence Probability versus Detection Probability

Axelrod, M

Lawrence Livermore National Laboratory

关键词: Probability;    Compliance;    Detection;    Sampling;    99 General And Miscellaneous//Mathematics, Computing, And Information Science;   

DOI  :  10.2172/875951
RP-ID  :  UCRL-TR-216207
RP-ID  :  W-7405-ENG-48
RP-ID  :  875951

美国|英语

来源: UNT Digital Library

【 摘 要 】

In a discovery sampling activity the auditor seeks to vet an inventory by measuring (or inspecting) a random sample of items from the inventory. When the auditor finds every sample item in compliance, he must then make a confidence statement about the whole inventory. For example, the auditor might say: ''We believe that this inventory of 100 items contains no more than 5 defectives with 95% confidence.'' Note this is a retrospective statement in that it asserts something about the inventory after the sample was selected and measured. Contrast this to the prospective statement: ''We will detect the existence of more than 5 defective items in this inventory with 95% probability.'' The former uses confidence probability while the latter uses detection probability. For a given sample size, the two probabilities need not be equal, indeed they could differ significantly. Both these probabilities critically depend on the auditor's prior belief about the number of defectives in the inventory and how he defines non-compliance. In other words, the answer strongly depends on how the question is framed.