STATISTICAL CONSIDERATIONS FOR LOT-BY-LOT ACCEPTANCE/REJECTION SAMPLING WITH AN ATTRIBUTE

Abstract

The standard procedure for lot-by-lot acceptance/rejection sampling described in literature makes always a decision. If there are more invalid items in the sample than specified, then the lot is rejected, otherwise the lot is accepted. However, if the actual proportion of invalid items in the lot is close to the accepted quality level then it is - in principle - impossible to control the customer's and the producer's risk with a reasonable sample size. The published standard procedure can result in acceptance of the lot even if there is insufficient information. We think this is unfair to the customer. Three statistical methods with a common acceptable quality limit are proposed and discussed: A simple procedure, confidence intervals, and finally a statistical test. The idea of a statistical test, its application in lot-by-lot acceptance/rejection sampling, and the proposed statistical test are described in detail and with examples. Additionally, determination of sample size and multiple step sampling plans are characterized.

Keywords

• confidence interval
• one-sided versus two-sided hypothesis
• sample size
• sampling plan
• statistical test