0

For example, we find 13 defects out of 1000 random samples tested. Now the issue is fixed. What is the minimum number of samples we need to test to confirm that the defect rate is reduced to 0.1% with 90% confidence? (We have no prior knowledge of this defect. We believe this issue happens randomly. We don't have a clear root cause and whether the issue is truly fixed. That's why we are relying on statistical observations to see whether the issue is fixed).

(I was pointed to the rule of 3 as a possible answer to this question. But, the rule of 3 seem to apply for when there are no defects and does not seem to use the information that we have a previous known failure rate.)

TreeCity
  • 1
  • 3
    Fixing the issue means confirming the defect rate is 0%, which you can't do (the next sample could always be defective, right?). Are you willing to amend your statement to something like "The defect rate is less than x" for some acceptable x? – Demetri Pananos Jan 25 '24 at 03:45
  • @DemetriPananos, thank you. I updated the question to reflect your comment. – TreeCity Jan 25 '24 at 08:20
  • Re the Rule of Three: that applies when there are no observed defects. With a minimum sample size, you certainly expect to see no defects, whence you can use the Rule of Three to compute that sample size. – whuber Jan 26 '24 at 13:21

0 Answers0