Dismiss Notice
You must be a registered member in order to post messages and view/download attached files in this forum.
Click here to register.

Overall Picture of Capability

Discussion in 'Other Quality and Business Related Topics' started by QualityKyle, May 28, 2019.

  1. QualityKyle

    QualityKyle New Member

    Joined:
    May 28, 2019
    Messages:
    1
    Likes Received:
    0
    Trophy Points:
    1
    I have a question regarding products with multiple testing variables. For example, a single product will be tested for pH, %solids and zeta potential. There are no rational population subgroups. Right now, I am calculating a Pp/Ppk among other metrics for all three tests separately. Is there any accepted methodology for looking at the product as a whole (e.g. averaging the Ppk from each test)? I assume not, and I know multivariate analysis is heavily debated as to its effectiveness. I guess I'm more interested in any practical approaches industry has taken in a similar position. Thank you for any input!
     
  2. Bev D

    Bev D Moderator Staff Member

    Joined:
    Jul 30, 2015
    Messages:
    419
    Likes Received:
    438
    Trophy Points:
    62
    Location:
    Maine
    Process capability indices are intended for single characteristics that are created by a unique “process”; hence the name process capability.
    Product capability cannot be described by a single “variables” (continuous data) distribution as the totality of characteristics com from different processes and will have very different distributions. You can describe the total capability as a defect rate. You can combine the rates using the rules of probability. Some defects will be independent and the conditional proababilty formula will apply. Others are not independent and you will need to use the appropriate probability formula to do this. OR you can simply calculate the actual total defect rate over a substantial period of time and report that. This latter way is the most accurate, easiest and most believable as it comes from actual results and not statistical estimation which is subject to errors of assumptions of which models might apply.
     

Share This Page