iSixSigma

IMR vs. Xbar-R

Six Sigma – iSixSigma Forums Old Forums General IMR vs. Xbar-R

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #52383

    Picklyk
    Participant

    Hello All,
    I am currently working on a project at work which involves a set of data grouped into subgroups of 3 and 10 of test data from the same sample coupon. Upon creating the Xbar-R charts for the n = 3 subgroup size and creating the Xbar-S charts for the n = 10 subgroup size, my colleague and I noticed a large number of points outside of the control limits on the Xbar charts, but most of the control limits on the S and R charts were within range. When we averaged the subgroup size, artifically making the sample size 1, and created an IMR chart, all of the data points were within the control limits on both the moving range chart and the individual value chart. My question is, is it acceptable to average the subgroup size and then construct IMR charts with this data?
    Thank you for your insight.
    Regards,
    Jay

    0
    #184268

    Old MBB
    Participant

    NO, it is not acceptable.  The control limits become tighter as sample size increases (to balance against the central limit theorem).  By plotting averages as individuals, you are hiding the presence of CLT.
    What you need to do is arrange your data in one column (in time order), and list the corresponding sample size in another column.  Most stats software packages (like Minitab) will allow you to generate an Xbar-R with varying n.  The contol limits will oscillate, but they will be correct.
    Good Luck

    0
    #184288

    Cone
    Participant

    Jay,Your book did not say you could do this. Don’t do it, it is dumb.

    0
    #184291

    Picklyk
    Participant

    Gary,
    The person I am working with on the project (the thermal spray process engineer) created IMR charts based on the data and preferred the IMR charts because they included more points within the control limits. Basically, I need to explain to him why we can’t do this.
    Jay

    0
    #184308

    BTDT
    Participant

    Jay:This data is unusual in some way I don’t understand. When data is non-normally distributed, there are more points beyond 3 s.d for the IMR chart than for the same data in subgroups using an X-barR chart.Do you want some help in digging deeper?Cheers, Alastair

    0
    #184311

    Sloan
    Participant

    Jay,
    Here’s the point I was trying to make in my other post.
    The purpose of a control chart isn’t to try to manipulate your data so that all of the data points fall within the control limits, the purpose is to tell you if your process contains only common variation or if it contains special cause variation. If your process contains special cause variation you are doing yourself (and your customer) a great disservice if you manipulate the data to make the out of control points disappear on the chart.
    Do the things mentioned here by BTDT, Old MBB and others to make sure you are treating your data correcly and once you are satisfied that the apparent variation is coming from your process and not your data, then go investigate the sources of your special cause variation and fix it. Manipulating the data to make your chart appear better is like sweeping dirt under the rug. The dirt didn’t go away you just hid it. That dirt could be toxic.

    0
Viewing 6 posts - 1 through 6 (of 6 total)

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