iSixSigma

Auditing

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • #49180

    Greenstein
    Participant

    I am rather new, as the question itself may reveal.  What is the percentage considered necessary in order to attach a sigma level to the population, when only a percentage can be audited?  In other words, IF sampling is the only manner in which the quality of output can be measured, what percentage of output must be audited in order to attach a specific sigma level to the result?

    0
    #167660

    MrMHead
    Participant

    It depends on how accurate you want/need to be.
    There is a “sample size calculator” to determine your degree of prescision: (excel format)
    For Continuous Data=SQRT(((1.96*StdDev)*(1.96*StdDev))/SampleSize)
    DiscreetData = SQRT((Power(1.96,2)*(PropDefect*(1-PropDefect)/SampleSize)))
    1.96 refers to the confidence interval of 0.05.
    I honestly don’t understand the formula completely.. or rather where it comes from.. but this would give you your margin of error.
    …hehe.. I just Googled Sample Size Calculator and came up with a bunch hits – they may help explain too.

    0
    #167661

    Greenstein
    Participant

    Thanks.  That is a good place to start.

    0
    #167663

    Fake ATI Alert
    Participant

    Why 1.96??

    0
    #167665

    MrMHead
    Participant

    >>>Why 1.96??
    Um … because that’s what the book said  ;-)
    It is the z-score for the end of the 95% (or 0.05) confidence interval.

    0
    #167668

    Fake ATI Alert
    Participant

    Calculating  Minimum Sample Size and  Sensitivity,Variable  Data:
    n=( Z*S/H)^2 to calculate  minimum sample  size  on  variable  data:
    n= minimum  sample  size
    Z=confidence  level.(when  in doubt use Z=1.96,per  the  following  tip.)
    S= the  population  sd
    h=the smallest change  we  want   to  be  able  to  sense
    (when  in  doubt,use  h=total tolerance/10 ,or h=0.6s)
    Good Luck

    0
    #168045

    AlexanderSO
    Participant

    Hi, sorry if I´m late to join the talk.  I suggest you use a MIL STD to definie your sample policy, the first step will be to define the type and level of inspection you want to perform and that depends on the requirements of your process, for example you choose a severe inspection when your process had shown bad performance according with the quality reqs, and a reduce inspection when the performance had been good (in the MIL STD 105D you can choose between severe, normal and reduced type), also you choose a level (I,II,III) and the selection allows to change the size of the sample without affecting the producer risk but changing the risk of the customer and the decision must be made knowing the effects and cost of each risk, in addition to this you must define an acceptance level for the production lot, the initial recomendation given in the literature is to use an acceptance of 95%, but it depends on the nature of your process.
    Over the selected sample you can perform quality and capacity analysis and estimate your sigma level and define action plans to achieve the desired level.

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

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