From: Stan Hilliard
Date: 22 Dec 1999
Time: 00:17:51
Greeting David,
Please explain more about the USP "zero tolerance" requirement that you need to comply with. Maybe I will be able to think of some suggestions for matching those requirements.
Meanwhile, concerning non-normality, variables that are not normally distributed can be dealt with by appling the appropriate normalizing transform to the raw data. The specific transform would depend on whether the raw data follows the log-normal, Poisson, or negative logrithmic. After transforming the data, the methods that apply to normally distributed variables can be used as in the tutorial. See:
www.samplingplans.com/modern3.htm
Another approach is to use a variables sampling plan for the mean, which will not require that the variable be normally distributed in the population. See more about sampling plans for the mean here:
www.samplingplans.com/outputvariablesmean.htm
The program TP414 will develop such sampling plans for the mean:
www.samplingplans.com/programtp414.htm
Sincerely, Stan Hilliard