I am working on a project for a company that is audit sampling its various products. Defects are of an attribute manner, so there are no variances within tolerance ranges to measure. For some of the products, the company is finding no defects out of as much as 50,000 pieces, and 1 defect out of as much as 100,000 pieces for others. Obviously, it would be expensive to keep sampling until five defects are found, just so a Normal approximation to the Binomial can be used. What is the appropriate method to calculate the confidence intervals or interval estimates for the defect rates?
Any help is very much appreciated.
--- Steve Landry firstname.lastname@example.org The Hartford Graduate Center