From: Stan Hilliard shilliard@samplingplans.com
Date: 17 Sep 2002
Time: 15:34:50
There is a method to determine sample size for variables data when you do not have any previous history or data. I will address the case where you will be measuring individual items. Of course, you must set specification limit(s) that reflect your standard as to when an item is nonconforming.
The method is this: Choose alpha, beta, AQL, and RQL exactly like you would for an attribute plan. You use these four sampling criteria to design a variables plan for fraction nonconforming for standard deviation unknown.
For this type of sampling plan, you calculate the sample average and sample standard deviation when the sample is finally taken. The fixed-n decision rule involves two numbers: n, K. The sample size is n and the decision limit K is compared to a calculated value from the sample. K_calculated=(Xbar-spec)/(S), where S=standard deviation and spec is a specification limit. You can have lower, upper, or both specification limits.
The software program TP414 will develop such a plan. With it, you can design both fixed-n and sequential sampling plans. www.samplingplans.com/programtp414.htm
The tutorial contains an example of this type of plan: www.samplingplans.com/modern3.htm