On Fri, 06 Dec 2002 16:04:50 +0100, Koen Vermeer wrote:
>>>I am not very familiar with the Anderson-Darling test. I only know it
>>>is an alternative to KS and it is not distribution free. The last thing
>>>was the mean reason I wasn't considering it for the moment. Maybe I'll
>>>have a look at it in th near future.
>> The Anderson-Darling test is as distribution free as the
>> Kolmogorov-Smirnov.
> Hmm, I thought I read somewhere that the KS test statistic is
> independent of the distribution that is tested against. Maybe I'm
> mistaken...
Just some experiments in Matlab: For n=10 and alpha=0.2, 0.15, 0.10, 0.05
respectively, I ran a Monte Carlo simulation of 100000 samples three
times. For a uniform distribution, I get
0.3202 0.3403 0.3657 0.4072
0.3202 0.3398 0.3659 0.4069
0.3201 0.3396 0.3657 0.4068
For a normal distribution, the results are
0.3199 0.3398 0.3659 0.4064
0.3205 0.3410 0.3665 0.4067
0.3201 0.3397 0.3652 0.4057
Looks quite similar. I haven't tried it for Anderson-Darling, but all my
sources say you need different critical values for different
distributions.
Koen
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
