Re: How to run Shapiro-Wilk test for each grouped

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Re: How to run Shapiro-Wilk test for each grouped

yvonnick noel
Iurie,
> Noel, thanks a lot. This will help me someday. But I have a question.
> When we run Shapiro-Wilk test, the homogenity of variances is a
> mandatory condition?
>  
No it is not. An homoscedasticity test only makes sense when you have a
grouping factor, and a normality test may of course be used in a variety
of contexts when you have a unique sample.

My point was: If you use gaussian models and assume homogeneity of
within-group variances, then testing normality is somewhat simplified,
since your model residuals are expected to be drawn from a unique normal
distribution, and only one normality test on the residuals is necessary
(no need for a loop).

Best wishes,

Yvonnick NOEL, PhD.
University of Brittany, Rennes 2
France

______________________________________________
[hidden email] 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.
Reply | Threaded
Open this post in threaded view
|

Re: How to run Shapiro-Wilk test for each grouped

Thomas Lumley
On Tue, 13 Apr 2010, Yvonnick Noel wrote:

> Iurie,
>> Noel, thanks a lot. This will help me someday. But I have a question.
>> When we run Shapiro-Wilk test, the homogenity of variances is a
>> mandatory condition?
>>  
> No it is not. An homoscedasticity test only makes sense when you have a
> grouping factor, and a normality test may of course be used in a variety of
> contexts when you have a unique sample.
>
> My point was: If you use gaussian models and assume homogeneity of
> within-group variances, then testing normality is somewhat simplified, since
> your model residuals are expected to be drawn from a unique normal
> distribution, and only one normality test on the residuals is necessary (no
> need for a loop).
>

Or, as some of us would say, *no* normality test on the residuals is necessary, which is even simpler.

     -thomas

Thomas Lumley Assoc. Professor, Biostatistics
[hidden email] University of Washington, Seattle

______________________________________________
[hidden email] 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.