For quality sampling, I need to ensure a defect rate of less than 5000 parts per million, which is 1/200, at a specific confidence level.
The goal is to have the smallest sample number for the confidence level. I wonder if a binomial test of significance is appropriate.
Say I sample 3000, and I only get 2 defective parts, then based on p=1/200 and n=3000, I sum the probability of getting 2 or less defective parts based on binomial distribution, if this probability is less than my significance level (say 0.05), then I conclude the defect rate is less than 5000 parts per million at said significance level.
Is my approach valid?