+ Reply to Thread
Results 1 to 4 of 4

Thread: Locating significantly different data points in a sample

  1. #1
    Points: 9, Level: 1
    Level completed: 17%, Points required for next Level: 41

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Locating significantly different data points in a sample




    Hi,

    I have question regarding some data I collected recently which can be represented as plain integers (basically the number of features (dots) counted on insects (the same species)) from 1 to 1000. Using R I concluded these data is log normally distributed. First does it make sense to ask the following question: " Which data points are significantly different from the background (mean)?" and second: how would one compute that ? which test would be appropriate for that ?

    am i making any sense ?

    thnx,
    mxs

  2. #2
    TS Contributor
    Points: 17,779, Level: 84
    Level completed: 86%, Points required for next Level: 71
    Karabiner's Avatar
    Location
    FC Schalke 04, Germany
    Posts
    2,542
    Thanks
    56
    Thanked 640 Times in 602 Posts

    Re: Locating significantly different data points in a sample

    I suppose you don't mean "significant" in the same sense
    that it is used within the significance testing framework
    What do you actually mean by "significant" here? E.g.
    are you looking for single outliers which might be due to
    measurement errors? Do you want to identify extreme
    cases and analyse them further?

    With kind regards

    K.

  3. #3
    Points: 9, Level: 1
    Level completed: 17%, Points required for next Level: 41

    Posts
    2
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: Locating significantly different data points in a sample

    Hi, and thnx for replying

    I am looking for a set of outliers that might be good candidates for further analysis. A set that might not belong to the group (my collected sample) Because since I am working on very similar species my hypothesis is that those that have a too many freckles are in fact members of the different subspecies and i want to separate them out

  4. #4
    TS Contributor
    Points: 17,779, Level: 84
    Level completed: 86%, Points required for next Level: 71
    Karabiner's Avatar
    Location
    FC Schalke 04, Germany
    Posts
    2,542
    Thanks
    56
    Thanked 640 Times in 602 Posts

    Re: Locating significantly different data points in a sample


    You could transform your log-normal distribution
    into normal and apply techniques developed
    for outlier detection in normal distributions,
    e.g. box-and-whisker plots, or formal tests
    such as those described here
    http://www.itl.nist.gov/div898/handb...on3/eda35h.htm

    Probably there are more suggestions if you
    search for outlier detection on the net.

    With kind regards

    K.

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats