iSixSigma

Use of Median for Non Normal Data

Six Sigma – iSixSigma › Forums › Old Forums › General › Use of Median for Non Normal Data

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • May 22, 2006 at 12:52 pm #43468

    dangwal
    ★ 10 Years ★
    Participant

    Hi All,
    Just was wandering on the usage of median of non normal data for baseline,which loses the focus of standard variation in baselining.Wanted to have some views on this.
    Thank you !!

    0
    May 22, 2006 at 2:32 pm #137909

    Brit
    ★ 10 Years ★
    Participant

    I tdepends on what you are measuring.  Non-normal data with a hard limit – like turnaround time tends to be better explained by the median.

    0
    May 22, 2006 at 5:28 pm #137914

    Elbrin
    ★ 10 Years ★
    Participant

    Brit,
     Median is used to describe central tendency in non-normal data, because of outliers.  A single extreme outlier can significantly skew the mean, but will have little effect on the median.  Variance, Range and Standard Deviation are still used as dispersion measures for non-normal data.  Non-parametric tests such as Mann-Whitney and Kurskal-Wallace rank the data points effectively creating a median for the test.  Non-parametric tests could be used on normal data; however they are not as precise as our other tests.

    0
    May 23, 2006 at 9:49 am #137956

    dangwal
    ★ 10 Years ★
    Participant

    Brit, its agreeable that median represents better in non normal data, but aren’t we losing the focus on the standard deviation, for example in cycle time projects i always find 20% of my data causing the non normal data.If i address this with a median i can show the improvements with the help of the median but always that 20% of my data will be causing a pinch to the customer, as i am losing the focus on standard deviation.
    I am just thinking of is thr any way of combing the median and standard deviation to measure the baseline and is a good representation of the same.

    0
    May 23, 2006 at 9:57 am #137958

    Rajesh Mohandas
    ★ 10 Years ★
    Participant

    I would suggest to convert the data to normal and do a capability analysis, depending on the spread or the Cp value you may further decide to proceed. Say your Cp value is 0.2 as of now depending on the target or USL and LSL and should focus for a goal of 1.33 and above. Discuss with your mentor and take a call in this regard.
    Regards,
    Rajesh.

    0
    May 23, 2006 at 10:07 am #137960

    dangwal
    ★ 10 Years ★
    Participant

    Guys, can we have the capacility analysis using median and std deviation as we have for mean and std deviation.????

    0
  • Author
    Posts
Viewing 6 posts - 1 through 6 (of 6 total)

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