iSixSigma

How to do DOE if the Y is attribute

Six Sigma – iSixSigma Forums Old Forums General How to do DOE if the Y is attribute

Viewing 18 posts - 1 through 18 (of 18 total)
  • Author
    Posts
  • #37455

    Ruddy
    Participant

    Is there anyone have the first-hand experience on DOE if the output is attribute data (pass/fail)?
    Usually we require the output should be variable. And I did read some articles to convert the attribute data to a percentage format. Is it realistic?
    My on-hand training materials all deal with variable output in DOE examples.

    0
    #110339

    Ted
    Member

    Michael – your question is even more complicated dependent upon the type X data you have. Looking to get a transfer function with binary output data and continuous inputs – you can look to Binary Logistic Regression (remember your transfer function is a log function). If the input data is also discrete the only way I have ever done it (and not more that I absolutely had to) was to grind it out. Going to your DOE structure (-1, +1) you build your conditions and then have to test each one for significance with the chi-square test. You can look at interaction this way as well but like I said it’s all by hand. Finally you have to interpret the significant factors relative to your pass/fail output. All in all not easy, not done often and always the bane of “transactional” BB.

    0
    #110379

    k.bhadrayya
    Participant

    Please send me your mail ID . I will send you a cse study where in attribute data on converting to continous dat has been used for DOE application.
     
    yours
    k.bhadrayya

    0
    #110381

    Tan Suok Ha
    Member

    Hi Bhadrayya,
    I happened to read your reply. I’m located at Malaysia Penang providing the DOE to the production also. Interested to know the example that you mentioned. Would you pls forward to me for reference. Thanks in advance.
    Yours sincerely, sh

    0
    #110392

    Y R Prasad
    Member

    Mr Bhadrayya,
                          I request you to pl send me  a copy of  the case to [email protected]. I am a six sigma black belt practicing in Hyderabad , India
     
    Thanks
    prasad
    1500 hrs
    Nov 8,2004

    0
    #110400

    Imtiyaz Biyabani
    Participant

    Mr Bhadrayya,
                          I request you to pl send me  a copy of  the case to my emailid  [email protected] . I am also interested in knowing the attribute DOE Case.
    Thanks in advance.
    Regards,
    Imtiyaz.

    0
    #110401

    ROSS
    Member

    Mr Bhadrayya,
                          I request you to pl send me  a copy of  the case to my emailid  tonsunhlj126.com . I am BB in china.
    Thanks in advance.
    Regards,
    Tony

    0
    #110404

    Vera
    Member

    Hi Mr. Bhadrayya:
    I am also interested in your case study. Would you please send it to me as well ([email protected])?
    Thanks
    Vera, SSBB
    USA

    0
    #110416

    Graeme Wilson
    Participant

    Have a look at:
    DOE With Attribute Response ()
    DOE With Attribute Response Data ()

    0
    #110422

    Habib Sdidiqui
    Participant

    For a defective output that is binomial (pass/fail), either a standard transformation with an arcsin square root function or a Freeman-Tukey transformation to stabilize the variance is a common practice in attribute DOE. A good reference is the 1994 work of Bisgaard and Fuller.

    0
    #110452

    Sandra Stokes
    Member

    I’m interested in a copy of your DOE for attribute data as well. Thanks in advance.

    0
    #110473

    Jonathon Andell
    Participant

    It’s possible, but it tends to be costly and slow.
    You need to run enough samples at each DOE setting so that the number of failures starts to approximate a continuous variable. For example, you easily can tell whether an experimental run produces 10 vs 100 defects. It’s harder to be sure the data are real if you try to distinguish between runs that create 45 vs 55 defects – at that point you might need to differentiate between 450 and 550 defects to be confident.
    You can perform the usual DOE ANOVA, and you also can consider some test of proportion and chi-square tests.
    Of course, to estimate how experimental settings impact variation in defect rtes you would need to replicate entire runs. That can get costly and time-consuming.

    0
    #110477

    Mike Smith
    Participant

    Mr. Bhadrayya
    I would alos like a copy
     
    Thanks
     
    Mike

    0
    #110480

    John Zhou
    Participant

    DoE Taguchi Methodology can assist you to conduct your Doe if Y is Attribute,
     
    Regards,
    John Zhou,

    0
    #110484

    Ruddy
    Participant

    Dear Bhadrayya,
    Your response is exciting. Appreciated if you can email me at [email protected]
     
    Thanks.
    Michael
     

    0
    #110488

    TF
    Member

    Dear Bhadrayya,
    Please send me the cse study where in attribute data on converting to continous dat has been used for DOE application.
    my e-mail is : [email protected]

    0
    #117967

    Johnny
    Participant

    Hello Bhadrayya
    I know this reply is a little late on this subject, I have only just researched this topic on the forum.
    I too have been struggling with Attribute Y for DOE’s.  I am involved with software and digital electronics, where the Y outputs are either correct/incorrect or on/off. 
    Please could you send me the case study so that I may have an example to learn from.
    Thankyou very much
    Johnny

    0
    #138400

    howe
    Participant

    May I also see the case study.  Please e-mail [email protected]

    0
Viewing 18 posts - 1 through 18 (of 18 total)

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