iSixSigma

DOE

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • #43134

    Sah
    Member

    We are trying to do a schedule a DOE today. Does anyone has a list of steps to follow?
    How does one determine which x’s are effecting the Y?
    Student

    0
    #136513

    Schuette
    Participant

    if p-value is less than .05 then your factor is significant.
    jimmm

    0
    #136514

    Sah
    Member

    sorry???/
    Can you elaborate/

    0
    #136517

    Robert Butler
    Participant

      Your post gives the impression that you just walked into work this morning and decided that today was a good day to schedule a DOE (whatever that is).  In the hopes that this isn’t the case I’ll give you the benefit of the doubt and try to offer some suggestions.
      If we assume you did all of the things you were supposed to do before getting around the thinking about a design then what you should have at this stage is the following: A list of variables of interest, a list of the individual ranges of the variables of interest, and a decision concerning how you want to proceed with investigating the variables (main effects only, mains + a few curvilinear effects, possible interactions, etc).  This will have driven your choice of design and you will have done your homework so you are already aware of how you will randomize the runs from the design over the course of the day.
      With all of this in hand and with the experiments run you are now in the position of examining the results. Since you say ” How does one determine which x’s are effecting the Y?” I will assume you are interested in identifying possible cause and effect relationships. 
      If this is the case then the usual procedure is to take the results of your design and run it through stepwise regression -both backward elimination and forward selection with replacement.  When you get the “final” models you will then run the usual regression diagnostics on the residuals to check for model adequacy, fit, influential data points, etc.
      Given that the model checks out to your satisfaction then you will take the model and use it to identify the settings of the X variables of interest that will result in optimum Y responses.  With this list in hand you will then go out on the line, set the X’s to the predicted levels and see if the resultant Y is within the prediction limits of the model.  You should run more than one confirmation run – using different X settings for different Y responses. If your results are in agreement with the model predictions you will be in the position of having an empirical equation that may be of value with respect to process control/improvement.

    0
    #136522

    TwoCents
    Member

    Don’t forget validation of any relevant measuring systems.  You could probably get that done real quick before you start. :).

    0
    #136537

    26and385
    Participant

    I can hear his brain exploding from here…..

    0
    #136539

    TwoCents
    Member

    Is that what that is?  I thought my neighbor was popping popcorn…
    We must’ve gotten his whole team :)

    0
Viewing 7 posts - 1 through 7 (of 7 total)

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