iSixSigma

To stage or not to stage, that is the question…

Six Sigma – iSixSigma Forums Old Forums General To stage or not to stage, that is the question…

Viewing 9 posts - 1 through 9 (of 9 total)
  • Author
    Posts
  • #49932

    Taylor
    Member

    At my current employer, we seem to have issue with staging for special cause detection and special cause resolution.  While I have always been taught in my almost 20 years of quality assurance and six sigma black belt work, that staging of a control chart (in our case, yield P charts) should be done when an out of control is investigated and the root cause is determined.  This isn’t the case at my current company.
    Since we are simply displaying yield, there seems to be a tendency to “pretty up” the charts by taking out points whether there are special causes identified.  These aren’t control charts that are meaningless because Minitab “squeezes” the control limits.  Rather, they show out of control points.
    So, do you as quality individuals believe that staging is the appropriate action in this case? I don’t.
     
     
     

    0
    #171422

    Taylor
    Participant

    In past experience, I would prefer to highlight the specail cause, other wise why are you going through the exercise.
     

    0
    #171519

    Rhineg
    Member

    Until you have identified and eliminated (with confirmation) a special cause, there is no reason to remove it from the population. Otherwise; as you indicated, someone is only drawing pretty pictures and is not improving the process, nor improving quality delivery to your customers.

    0
    #171620

    Taylor
    Member

    I want to thank both of you for your responses and I would encourage anyone else to provide their input.  So far the best definition seems to be that of Rhineg.  If you have a better definition, I’m listening.

    0
    #171623

    Outlier, MDSB
    Participant

    I’m not sure what you mean by “staging.” Do you mean resetting calculation parameters from that out of control point so that Minitab draws a vertical line through the chart and give you a new p-bar? I’m sure it is just a terminology thing for me.

    0
    #171634

    Taylor
    Member

    Minitab recalculates UCL, LCL and average p for every data point added/subtracted/modified in the population automatically (unless programmed otherwise).  Staging is meant to indicate a major “change” to the process.  In our situation, it means that a special cause has been identified and “handled”.  Once done, this point is taken out of the population. 
    At issue is the lack of “closure” on the special causes and that removing the point(s) somehow makes the control chart more meaningful.
    Mechanically, your description is correct, the logic and philosophy of removing or staging the data is what is in question.  Thanks!

    0
    #171641

    Marlon Brando
    Participant

    No
    The real question is to be or not to be?

    0
    #171643

    Outlier, MDSB
    Participant

    Todd,
    Thanks for the clarification, that’s what I thought you meant.
    I’m not in favor of removing data points form the chart even if you have identified a special cause for an out of control situation. At most, I might flag the data point and tell Mini to omit that point from the calculation of the center line and control limits. That way, my control limits remain a reflection of a stable process and don’t artificially move the control limits, but it also doesn’t hide the fact that we had an event of some kind. I will do this only after I have identified the special cause as such. I will NEVER reomve an outlier simply to make the chart look better.
    I only like to freeze my historic process and recalculate new centerline and limits when I know that my process has really changed.
    On a related topic, I’ve noticed that control charting in “modern” times (using a coupter and minitab) has probably strayed a bit from the way Shewhart intended them work. Before the days of computers when control charts were calculated and drawn by hand, you would take a sample of data, calculate your centerline and UCL/LCL then plot your chart. From then on (until you had a major process change) with every new data point added, you simply added the data point to the existing chart. Your centerline, UCL and LCL remained static. However with Minitab, the tendancy is to add a datpoint to the worksheet and recalculate the entire control chart with all of the data including the new data. Essentially, you establish a new baseline every time you add a data point. Your centerline and CL’s are constantly shifting.
    My concern about it is that your constantly changing control chart may be less robust to detect a shift in the process. I haven’t done any kind of formal analysis on this but I wonder if others have noticed the same distinction between control charts with a computer VS control charts drawn by hand. Does the subtle constant shifting have an effect on the chart’s ability to detect special cause?
     

    0
    #171699

    KKN
    Participant

    The way you just described control limits is the way I changed our control charting when I joined my new job.  We set the data points minitab uses to calculate the CL’s to an  established “stable” section of time.  We only change them when we have seen a change in the process (not special causes) and the change has stablized.  We change whether or not the process change is good or bad. 

    0
Viewing 9 posts - 1 through 9 (of 9 total)

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