i have 3 group variables (faculty at the uni, workcategory and careertype). faculty has i think 5 undercategories, workcategory 8 and careertype has 9

and for my hypothesis testing i need 13 Variables (e.g. grade, fantasy, self-efficacy, ...)

do i need to test for normality each group variable on each level for each variable? or is it sufficient to just test each groupvariable for normality and then the distribution of each variable?

  • 1
    $\begingroup$ What for do you want to test it? Basically, if you want to assume overall normality, test overall normality, if you want to assume within group normality, test against it... $\endgroup$
    – Tim
    Commented May 9, 2017 at 7:42
  • $\begingroup$ it's pre testing for parametric testing and i want to see if for example fantasy is diffrent in diffrent kinde of faculties --> do i need then to test for normality in the group faculty for the variable fantasy? $\endgroup$
    – Kerstin
    Commented May 9, 2017 at 8:05
  • 1
    $\begingroup$ orthodox capitalisation makes posts easier for many to read. it may seem cute not to use the shift key, but it's still non-standard. $\endgroup$
    – Nick Cox
    Commented May 9, 2017 at 8:13
  • 1
    $\begingroup$ No trolling. Just editing your post and making other suggestions for improving it. That's not trolling. $\endgroup$
    – Nick Cox
    Commented May 9, 2017 at 8:48
  • 1
    $\begingroup$ @Kerstin Please adopt a professional tone, especially when people are trying to help you by explaining the norms here. Rudeness like that (accusing someone of trolling and deliberately inserting stray capitals into your comment) is not in keeping with stackexchange policy. You may not originally be English speaking but that doesn't mean that things that make your post hard to read should be left alone. Please take the effort to try to improve your post as was suggested. $\endgroup$
    – Glen_b
    Commented May 9, 2017 at 8:54

1 Answer 1


If you are going to use ordinary least squares regression, then you don't need to test for normality of the dependent variable at all. OLS regression does not make assumptions about the distribution of the dependent variable, it makes assumptions about the distribution of the errors, as estimated by the residuals.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.