[R] Normal distribution test

[Markus Mehrwald]

Thank you for all the answers!

Kjetil,

I am not sure if we are talking about the same thing. I only have a two 
dimensional normal distribution which leads to three dimensional data. 
You mean with "reject" I should not do such a test?
My data files contain about 260000 points which I can reduce to the 
half. The data is created through a sum of two dim. Gaussian profiles 
(or just one). It is easy to fit a 2 dim. Gaussian function but this 
does not take material properties into account so I cannot be sure that 
the points are realy normal distributed. What I want to do with that is 
to proof that the model (at least one 2D Gaussian function) is working 
correctly or if I have to think of a different one.

Regards,
Markus


Kjetil Halvorsen schrieb:
> On Tue, Nov 17, 2009 at 11:17 AM, Markus Mehrwald <mehrwald at ira.uka.de> wrote:
>> Hi all,
>>
>> I am completely new to R and my knowledge of statistics is quite small so I
>> hope you can help my.
>> I have three dimensional point data which represents (and this is what I do
>> not know for sure) a normal distribution. Now I want to test if this is true
> 
> I suppose you want to say you have a sample of three-dim data, say
> represented be vectors x1,x2,x3,
> and your question is if this data (x1|_1,x2_1, x3_1),...,(x1_n,x2_n, x3_n)
> are generated by a three-dim multinormal distribution. That is very
> simple, a very good
> test is to simply say "reject".  I have never seen three-dim data
> which are truly
> multinormal.  So a better question is to ask if amultinormal
> distribution can be an acceptable
> approximation, but then we need to know what is your purpose of
> analysis! If you are interested in
> extremes or extrere quantiles, then a normal approx is never safe.
> 
> If you want a statistical test, then a multivariate extension of
> shapiro-wilk is in
> install.packages("mvnormtest", dep=TRUE)
> library(mvnormtest)
> ?mshapiro.test
> 
> kjetil
> 
> 
>> or not and as I can remember from statistics lessons I can use Chi-Square
>> test for distribution test. BUT: I have realy no idea how to do this with R
>> and additionally if my assumptions are correct and if this is possible with
>> R at all.
>>
>> Thank you very much in advance for any answer.
>> Markus
>>
>> ______________________________________________
>> R-help at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>>
> 

-- 
Dipl.-Inform. Med. Markus Mehrwald
Institut für Prozessrechentechnik, Automation und Robotik
Medizin-Gruppe
Universität Karlsruhe (TH)
Gebäude 40.28, Zimmer 110
Engler-Bunte-Ring 8
76131 Karlsruhe

Fon: +49 (721) 608-7113
Fax: +49 (721) 608-7141