From: Stan Hilliard
Date: 02 Apr 2000
Time: 23:12:27
Greetings Dennis,
The pervasive errors are qualitative, so you need an attribute sampling plan. You can choose between two types of plan - audit (estimation) sampling or acceptance sampling depending on whether you want to specify the precision of the sample estimate or to specify the acceptable and rejectable quality levels.
> How large a sample should be used?
1) Audit sampling calculates the sample size that will provide a specified margin of error of the sample estimate of the population quality level. Stated differently, audit sampling imposes a level of precision on the measurement.
2) Acceptance sampling calculates a decision rule consisting of both the sample size and decision limit. The decision rule will accept or reject the quality levels that you specify. The decision rule has both producers and consumers risks.
The software to support these two objectives is here:
www.samplingplans.com/software.htm
> Are there references that > justify this sample size?
The numerical sampling requirement that you choose (in 1 or 2 above) determines the sample size. As you try to make this requirement more aggressive, the sample size goes up. High sample size requires more time to audit, so your workload is tied directly to the sampling requirement. You arrive at a plan by striking a balance between these factors.
Sincerely, Stan Hilliard