Re: The search of no-normality

Kevin E. Thorpe wrote:
Herman Rubin wrote:
In article <e7oi0m$kch$1@xxxxxxxxxxxxxxxx>,
Jeremy Miles <jnvm1@xxxxxxxxxx> wrote:
"Luis A. Afonso" wrote:
Against the obsessive persecutors of no-normality
Suppose that the K-S test fails to reject H0 at the 5% level (the
Distribution is normal).

No. We conclude that we do not have evidence to reject the null
hypothesis that the distribution was drawn from a population with a
normal distribution. Subtly different, but it's important.
e.g. x = (1, 2, 3, 4, 5, 6, 7, 8, 9, 10).
K-S test p-value is 1.00. So you think that's a normal distribution?
You must have done something wrong. The K-S test p-value
is almost never 1; it may be close, but I do not think that
this sample is that close.

I got p = 0.9999 when I compared that sample to
a normal with mean 5.5 and sd of 3.027650
(the sample mean and sd).


SPSS gives p = .999989 for an asymptotic test, and p = .999858 for an exact test (both 2-tailed).

--
Bruce Weaver
bweaver@xxxxxxxxxxxx
www.angelfire.com/wv/bwhomedir
.

Relevant Pages

  • Distribution Fitting and GoF Tests
    ... The distribution of the data decays in an exponential ... for 1000 random subsets) and observe the average p-value, ... which they determined the better fit. ... I wrote a script in MATLAB to see the results the K-S test ...
    (comp.soft-sys.matlab)
  • Re: Distribution Fitting and GoF Tests
    ... with that much data you're very likely to find any distribution test ... It's true that the K-S test and some others require that the parameters be ... Estimating them tends to make the p-value too big. ... Your results seem to indicate that the p-value is fairly accurate. ...
    (comp.soft-sys.matlab)
  • Re: Parametric Bootstrapped Kolmogorov-Smirnov GoF test
    ... initial data values and the theoretical distribution I've read I have ... Thus using a fitted distribution in the GOF test might be regarded as introducing a "bias". ... sample K-S test) giving say ks.bstats ...
    (sci.stat.edu)
  • Re: The search of no-normality
    ... "Luis A. Afonso" wrote: ... Suppose that the K-S test fails to reject H0 at the 5% level (the ... We conclude that we do not have evidence to reject the null hypothesis that the distribution was drawn from a population with a normal distribution. ... K-S test p-value is 1.00. ...
    (sci.stat.math)
  • Re: how to get confidence interval from Monte Carlo simulations?
    ... K-S test is constructed to compare data with a Normal ... Distribution that is DEFINED at the start and is ... useless for a single unknown sample. ...
    (sci.stat.math)