Pass/Fail Sample Data to Determine Sigma Level

Six Sigma – iSixSigma Forums General Forums General Pass/Fail Sample Data to Determine Sigma Level

Viewing 4 posts - 1 through 4 (of 4 total)
  • Author
  • #53689


    My process appears to be very stable. Missing documents are a key customer measurement, and at a critical process we are sampling by counting the number of records in the box (actual) and comparing it to the count generated by our processing machine. We have taken numerous samples on a daily weekly basis.

    In processing millions of records per year, we see practically zero defects in the thousands of records we’ve sampled. How would I go about using this data to determine sigma level? Is there a way to convert pass/fail to Yield and Yield to Sigma level.

    How many samples would i have to take to achieve a certain confidence level? (99%) assuming well over 1,000,000 are processed each year?

    Any experience with this issue or guidance is appreciated.



    You need to calculate DPU and then calculate sigma level from DPU through MINITAB.Again sample size can be calculated from MINITAB



    Unfortunately, the way you pose your question is a bit vague to properly answer:

    First, it appears that you are dealing with continuous data (number of records in a box), but then switch to discussing discrete data (number of records, which will require some kind of discrete calculation; otherwise how will you track a missing record ?). Nail down how you want track your data (hint:continuous) before you take the next step.

    Second, you discuss wanting to find the number of samples required to achieve a certain confidence interval, but nothing about the acceptable error (+/-) that you are willing to accept. To get to a sample size, you’ll need to determine the acceptable error as well.

    In this case, since it appears that you already have pulled your sample sets, I’d let the data tell me what the CI and error are based upon the number of samples you have. If the error is unacceptably large, then you’ll need to capture more data.

    Finally, I’ll agree with aravindags, Minitab, or your stats tool of choice, is best for calculating your CI, error, and sigma values.




    Missing documents seems to be a very naiive measurement (no offense). What about the data that the documents contain? Missing would only seem to be one type of potential error. Missing data, incorrect data, mis-entered data, etc. would also seem to be critical errors that would be important.

Viewing 4 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic.