iSixSigma

Dealing with non-normal distributions

Six Sigma – iSixSigma Forums Old Forums General Dealing with non-normal distributions

This topic contains 12 replies, has 6 voices, and was last updated by  annon 14 years, 3 months ago.

Viewing 13 posts - 1 through 13 (of 13 total)
  • Author
    Posts
  • #40159

    annon
    Participant

    Is the Empirical Rule applicable to non-normal distributions?
    Once non-normality is determined, how does one then determine what type of distribution he or she is dealing with? 
    Thanks.
     

    0
    #123765

    Darth
    Participant

    1.  No.  But, Chebychev applies a similar concept but to any continuous distribution.  Research it.
    2.  Fit a model, usually a probability plot and look for the p value.  Easy to do with Minitab.

    0
    #123766

    annon
    Participant

    Thanks Darth.

    0
    #123771

    Dayton
    Member

    Can the empirical rule be applied to non-normal distributions?   I’m also tempted to just say No.  
     
    But you need to ask the question could a non-normal distribution plot both symmetrically and bell-shaped?   The answer to that is Yes.  
     
    Therefore could you conceivably find a non-normal but symmetrical bell-shaped plot that met the empirical rule segmentation ratio?   Maybe.
     
    But rather than futz with it why would you want to try to apply the empirical rule, if you knew you were dealing with non-normality?   Why potentially compound your margin of error by misapplying tool fitments?
     Vinny

    0
    #123772

    annon
    Participant

    FYI
    Chebychev and Empirical Rules
    Knowing the mean and standard deviation of a sample or a population gives us a good idea of where most of the data values are because of the following two rules:

    ‘s Rule The proportion of observations within k standard deviations of the mean, where , is at least , i.e., at least 75%, 89%, and 94% of the data are within 2, 3, and 4 standard deviations of the mean, respectively.
    Empirical Rule If data follow a bell-shaped curve, then approximately 68%, 95%, and 99.7% of the data are within 1, 2, and 3 standard deviations of the mean, respectively.
    EXAMPLE: A pharmaceutical company manufactures vitamin pills which contain an average of 507 grams of vitamin C with a standard deviation of 3 grams. Using Chebychev’s rule, we know that at least

    or 75% of the vitamin pills are within k=2 standard deviations of the mean. That is, at least 75% of the vitamin pills will have between 501 and 513 grams of vitamin C, i.e.,

    EXAMPLE: If the distribution of vitamin C amounts in the previous example is bell shaped, then we can get even more precise results by using the empirical rule. Under these conditions, approximately 68% of the vitamin pills have a vitamin C content in the interval [507-3,507+3]=[504,510], 95% are in the interval [507-2(3),507+2(3)]=[501,513], and 99.7% are in the interval [507-3(3),507+3(3)]=[498,516].
    NOTE: Chebychev’s rule gives only a minimum proportion of observations which lie within k standard deviations of the mean.

    0
    #123774

    annon
    Participant

    Just looking to broaden my understanding.  Can someone indicate where in minitab you would go to deternine the the type of non-normal distribution? Thanks.  

    0
    #123777

    annon
    Participant

    Disregard.  Thanks.

    0
    #123778

    Dayton
    Member

    Not knowing what you are trying to do, I suppose I’d use a goodness of fit test – either the Anderson-Darling or the Kolmogorov-Smirnov test depending on sensitivity needed and how much weight needed to be given to the tails of the distribution. 
     
    Vinny
      

    0
    #123782

    “Ken”
    Participant

    Vinny,
    You’re on the mark with your response.  The E-rule applies to symmetric bell-shaped distributions providing approximate probabilities between the mean and xSDs from the mean.  Good call!
    Ken

    0
    #123786

    Darth
    Participant

    Try Graph/Probability Plot……try a few different distributions and see if one fits the data better than another.  The p value should be of some help.

    0
    #123794

    BTDT
    Participant

    “Distribution ID plot” will run the tests for about 11 different distributions all in one whack. Also look at https://www.isixsigma.com/library/content/c050725a.asp this week.BTDT

    0
    #123841

    FTSBB
    Participant

    Darth and I have argued about this in the past, but I’m still a liberal when it comes to the empirical rule.  For example, I fully agree that a perfect normal distribution will have 99.73% of the points between +/- 3 sigma.  However, distributions such as a perfect uniform and right triangle distribution will have 100% within +/- 3 sigma, so they don’t viloate the rule; very “heavy tailed” distributions like Chi-square w/ 2 degrees of freedom and exponential will NEVER have points below -3 sigma and only about 2-3% over the +3 sigma level.  To answer your question: no, they don’t fit the rigid statistical rule, but they sure do come close…

    0
    #123844

    annon
    Participant

    Thanks to everyone.  Very helpful.

    0
Viewing 13 posts - 1 through 13 (of 13 total)

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