+ Reply to Thread
Results 1 to 4 of 4

Thread: Normal versus non-normal - presentation of results in publication

  1. #1
    Points: 2,062, Level: 27
    Level completed: 42%, Points required for next Level: 88

    Posts
    17
    Thanks
    5
    Thanked 0 Times in 0 Posts

    Normal versus non-normal - presentation of results in publication




    Dear all

    Can anyone help me with the following question:
    I have a outcome variable that is compared for different grouping variables. For some of the grouping variables the outcome looks in all groups normally distributed and hence I would go with parametric test. But there is one variable, if I take that one, the distribution in the corresponding groups does not look normal anymore. I wonder if in a publication for the same variables I should be consistent with the analysis method or if it is ok if once I use non-parametric and another time parametric approach.
    As an example: Let the outcome be blood pressure. Blood pressure for females and males would be normally distributed => To test for difference between gender a t-test could be used. Blood pressure according to a certain co-morbidity (Having the co-morbity yes versus no) does not look normally distributed in the groups. Should I use for the same outcome (blood pressure) different methods or be consistent and always use the same method for the outcome blood pressure.

    Thanks in advance for your help.

    Lisa

  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: Normal versus non-normal - presentation of results in publication

    Whether utilization of a "nonparametric" test is recommended, depends
    largely on the sample size, which unfortunately was not mentioned here.
    The usual approaches from the general linear model, such als ANOVA
    or t-test or multiple regression analysis work quite fine with "non-normal"
    distributions within groups, or of the residual, respectively, if sample size
    is large enough (the central limit theorem probably applies already with
    n > 30).

    With kind regards

    K.

  3. #3
    Points: 2,062, Level: 27
    Level completed: 42%, Points required for next Level: 88

    Posts
    17
    Thanks
    5
    Thanked 0 Times in 0 Posts

    Re: Normal versus non-normal - presentation of results in publication

    Thanks for your answer.
    Yes, that's true that with large sample size it should not be a problem.
    But what would recommend in case of a small sample size? Or in another case where we want to test for equality of variances, where the violation of the normality would be problematic for the F-Test (and the same situation in some groups the variable is normal in others not).

    Best regards
    Lisa

  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: Normal versus non-normal - presentation of results in publication


    But what would recommend in case of a small sample size?
    Why don't you just tell us the figure?
    Or in another case where we want to test for equality of variances, where the violation of the normality would be problematic for the F-Test (and the same situation in some groups the variable is normal in others not).
    http://en.wikipedia.org/wiki/F-test_...ces#Properties

    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