iSixSigma

How to measure variance with small volume of data

Six Sigma – iSixSigma › Forums › Old Forums › General › How to measure variance with small volume of data

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • September 26, 2006 at 2:33 pm #44705

    PWR2006
    ★ 10 Years ★
    Participant

    What is the best way to set the control limits with only 13 points of rolling data ?If I try and use I-MR charting then the 3sd +/- varies wildly with any exceptional single point changes.

    0
    September 27, 2006 at 7:19 am #143873

    Mishra
    ★ 10 Years ★
    Member

    Hi,
    Control limits u will not set up.. Control limit are dervied from the data..
    Secondly for Stastically you should have minimum 21 data points… You can still plot control limits with 13 data points however, u still need wait till 21 data points to study the process capability. However u can study with the 13 data points on variation. I.e Spls cause or a Common cause
     

    0
    September 27, 2006 at 8:18 am #143875

    Orang_Utan
    ★ 10 Years ★
    Participant

    Why must be more than 21 points? Why not 10 points or 50 points or others?Any logic behind your conjecture?

    0
    October 2, 2006 at 12:12 pm #144155

    Kapik
    ★ 10 Years ★
    Participant

    In Wheeler’s “Guide to Data Analysis…”, he discusses the number of data required create good control limits for a baseline.  He used simulation to show that as few as 4 or 5 points can be used with high level of confidence that an out-of-control point is not a false alarm.  More is better, but one should not hold off on control charting just because of only a few data points.

    0
    October 2, 2006 at 2:52 pm #144171

    melvin
    ★ 10 Years ★
    Participant

    Step back and consider what you are trying to accomplish during this initial period of establishing a control chart.  Is it to develop enough understanding of the process so that you can develop a filter to separate common cause from special cause process behaviour?
    If so, the initial points are about observing the process and start to develop that filter.  I’d have to believe that a point during these early stages that causes the control limits to fluctuate widely would be worth investigating as a potential special cause. 
    If they are due to special causes – should they be included?
    If they are no due to special cause – use them to set control limits?
    Bob 

    0
    October 2, 2006 at 3:48 pm #144175

    Savage
    ★ 10 Years ★
    Participant

    Are you sure he didn’t mean subgroup sizes of 4 or 5??? 

    0
    October 2, 2006 at 5:46 pm #144181

    MCK
    ★ 20 Years ★
    Participant

    I attended Dr. Wheeler’s classes on the Guide to Data Analysis (highly recommended) and he did really mean that you can start plotting a control chart with as few as 4 data points.  Believe it or not, the limits do not vary that much as you add more data, but as suggested in an earlier response, it pays to research the high and low points for special cause that will help to continually reduce the process variation.
    Good Luck!

    0
  • Author
    Posts
Viewing 7 posts - 1 through 7 (of 7 total)

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