iSixSigma

DOE CI?

This topic contains 4 replies, has 4 voices, and was last updated by  annon 14 years ago.

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • #41293

    annon
    Participant

    Do DOEs have CI or probabilities associated with them?  How do you quantify the probability of success?

    0
    #129332

    Robert Butler
    Participant

    Let’s think about the definition of the word “experiment”. It can be viewed as: test, trial, a tentative procedure or policy, an operation carried out under controlled conditions in order to discover an unknown effect or law. By definition such a process will not have a confidence interval or probability of success. If it were possible to do such a thing there wouldn’t be much need for any kind of experimentation. All you would have to do would be – build a design, run it through your POS (Probability of Success) algorithm, pick the best experiment, run it, and make tons of money.A good design needs to have both good and bad data – that is data with both good and bad responses. Consequently, what you should do is the following: After you have built your design (and here we are assuming you did all of the things you should have done in order to arrive at your final list of variables) sit down with your group and ask them to identify those design points whose experimental conditions will most likely result in a bad response (i.e. scrap, re-work, etc.) If you have done a good job picking your team you will have people on board whose prior experience with the process will allow them to do this. This exercise has two benefits:1. You will have checked to make sure your final design isn’t looking at a region where most of the effort will result in experiments of little value.2. You will have a “worst case” estimate of the amount of scrap your experimental effort is likely to generate. Often, I’ve found #2 to be very valuable when dealing with people who view the DOE as a waste of time. If you contrast the projected dollar loss from your DOE due to the scrap with the hoped for dollar gain should you find an improvement you can often win over people who heretofore had been either excessively neutral or perhaps even hostile to your ideas.I had a discussion concerning this issue with an earlier poster. The links to the discussion may be of some help.https://www.isixsigma.com/forum/showmessage.asp?messageID=54976https://www.isixsigma.com/forum/showmessage.asp?messageID=55014

    0
    #129333

    Mr IAM
    Participant

    Hmm…  This is a strange question.   DoE’s are for determining what factors and at what levels have an effect on some output.  “Success” is not a term I have heard used in the context of designed experiments.  I think every experiment is a success…. 
    What does success mean to you in terms of a designed experiment?

    0
    #129339

    Craig
    Participant

    DOES  have P-values associated with the ANOVA model and the effects estimates.
    You can do confidence interval calculations as well. As an example, you might want to compare differences in treatment means
    m 1 – m2
    There is a method for calculating at 95% confidence interval on this difference based on your DOE results. See “Statistical Methods and Data Analysis” fifth edition by Ott and Longnecker (page 917)
     
    Good luck!

    0
    #129342

    annon
    Participant

    Great…thanks!

    0
Viewing 5 posts - 1 through 5 (of 5 total)

The forum ‘General’ is closed to new topics and replies.