[ | Next | Previous | Up ]

Re: Risks

From: Stan Hilliard
Date: 15 Dec 2001
Time: 16:00:59

Comments

Technically, a sampling plan is an application of statistical hypothesis testing.

Here is the reasoning -- neglecting such issues as one-sided versus two-sided, upper versus lower limits, etc.

With AQL_mean as the null hypothesis, when a sample average (Xbar) falls in the critical (rejection) region, you can reject the hypothesis that the population mean m' is at the AQL_mean. Therefore you reject the lot.

With RQL_mean as the null hypothesis, when a sample average (Xbar) falls in the critical (rejection) region you can reject the hypothesis that the population mean m' is at the RQL_mean. Therefore you accept the lot.

The above two paragraphs contain the logic used by the equations that calculate the sample size and decision limit(s). However, in everyday casual discussions it is often best to avoid this explanation because it has a "double negative" aspect and requires a lot of concentration to sink in. In other words, it screws up most people.

A more straightforward way to explain the protection provided by a sampling plan is with its oc curve. The oc curve approach is described here: www.samplingplans.com/usingoccurves.htm

Stan Hilliard


Last changed: November 20, 2007