Re: Contradictory tests of normality

On Sun, 28 Dec 2008 19:14:11 -0000, "Dan" <tesfes@xxxxxxxxx> wrote:

I'm proper stumped on a problem in trying to decide whether a distribution
is normal or not. A histogram shows virtually no left tail, and therefore
positive skew and negative kurtosis. But Stata gives skew and kurtosis
values of .21 and 2.34 - quite close to 0 and 3, as they should be for a
normal distribution. Then on the other hand, Stata's test of normality
rejects normality, giving a probability of 0.0000. So why are the skew and
kurtosis values so close to zero and three?

I can't work this out and would really appreciate some help


SPSS uses a convention that the normal values of
skewness and kurtosis are both centered at zero.
Maybe Stata does, too? You could check - by documentation
or test data.

By what I've usually seen, -0.66 would be a moderately
large amount of kurtosis. I'm usually more worried
about skew, and outliers in that tail that you do have.

Again, as Ray says, if your sample is large enough,
any sample will reject when one tail is missing.

--
Rich Ulrich
.

Relevant Pages

  • Re: Contradictory tests of normality
    ... positive skew and negative kurtosis. ... rejects normality, ... as being close to zero and three. ...
    (sci.stat.consult)
  • Re: Contradictory tests of normality
    ... positive skew and negative kurtosis. ... rejects normality, ... distribution. ...
    (sci.stat.consult)
  • Re: Contradictory tests of normality
    ... positive skew and negative kurtosis. ... rejects normality, ... I am a student social scientist so forgive me if I sound amateurish, but I always thought skew/kurtosis demonstrate a normal distribution for the sample because skew measures symmetry and kurtosis measures the height of the middle values. ...
    (sci.stat.consult)
  • Re: Contradictory tests of normality
    ... positive skew and negative kurtosis. ... rejects normality, ... distribution. ...
    (sci.stat.consult)
  • RE: KURT and SKEW functions
    ... values for testing normality with the sample skewness and sample kurtosis. ... we should get a nearly close normal distribution as we ...
    (microsoft.public.excel.misc)
Loading