From: Stan Hilliard
Date: 4/16/2004
Time: 12:07:50 PM
Greetings again Stan,
YOU SAID: "The ANSI stds and sampling plan programs for variables address normally distributed properties. Assume the property I am sampling for is not normally distributed."
A method that I have used with non-normal distributions is this.
1) Use historical data from several lots.
2) Subtract from each measurement its lot average. (Now the lot average of these transformed points is zero, but the within-lot standard deviation is unchanged.)
3) Combined the data from all lots into one group. It will have a mean of zero and a standard deviation equal to the pooled within-lot standard deviation.
4) Plot this transformed data on normal probability paper.
5) From the normal chart, pick off two standard deviations -- one for the upper side of the distribution and one for the lower side.
6) Create two one-sided variables sampling plans, using the lower SD for the lower ISL plan and the upper SD for the upper ISL plan. Use the same sample size for each for simplicity. This will occur automatically if you use the same alpha, beta, AQL, RQL for the two plans.
7) The two plans can be either percent-nonconforming plans or plans for the mean, where the mean where, for each plan (lower and upper) the AQL-mean and the RQL-mean are calculated from the respective ISL and within-lot SD.
8) The two acceptance numbers define the acceptance range of the sample average, taking into account non-normality.
The software program TP414 will develop the the two individual variables sampling plans. They can be either fixed-n or sequential plans, ISL plans (fraction nonconforming) or mean plans.
www.samplingplans.com/programtp414.htm
Stan Hilliard.