iSixSigma

Appropriate Control Chart

Six Sigma – iSixSigma Forums Old Forums General Appropriate Control Chart

Viewing 3 posts - 1 through 3 (of 3 total)
  • Author
    Posts
  • #35356

    Nwajei
    Participant

    To my six sigma friends…
    I need to apply a control chart to a call rate metric.  The call rate metric is an annualized rate of calls to our service centers as a function of the installed population.  (So a 2.0 rate would mean that the average installed machine would require 2.0 calls per year or a call every six months). As the install population grows (varying group sizes) and there is no limit to the calls (theoretically), I could assume that a U chart is most appropriate. Yet, I’m not sure if this is representative of a Poisson distribution.  So instead I have used the regular I-chart to illustrate the process control profile.  Am I correct?
    Any thoughts would be appreciated.
    Thank you.
    Frank

    0
    #100325

    PhilWhateley
    Participant

    Frank
    Traditional attributes charts (p, np, c, u) all rely on a theoretical model (binomial distribution or Poisson distribution) to determine the standard deviation from the mean. However, this is ONLY valid when the requirements of the underlying theoretical model are satisfied.
     
    When the requirements of the theoretical model are not valid, the only way to determine the standard deviation is to do as you have done – use an individuals chart and use the average moving range as an EMPIRICAL measure of the standard deviation. This will always give the right answer, whether or not the requirements of the binomial or Poisson distribution are satisfied (The only general advantage of the traditional attributes chart is therefore increased degrees of freedom)
    Don Wheeler’s book “Advanced Topics in Statistical Process Control” is a good source on this, (including information on the theoretical requirements.)

    0
    #100581

    ALEK DE
    Participant

    Under a confusing scenario best decision is to use the I-MR  atleast I think . So what you have done looks okay . As regard this particular scenarion is considered how are you drawing the U chart ? Are you taking no of installations at different places as different sub group sizes & then count of calls from  each of these groups ? If yes , then U chart should  correctly serve your purpose (identifying the common & special cause variation)..
    Thanks
    Alek

    0
Viewing 3 posts - 1 through 3 (of 3 total)

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