R help - Sep 2010 - kstest vs shapirotest

Dear R-users

Idea: 
Analysing tree height frequency with hist(), normal distribution (ks.test &
shapiro.test) and skewness (package e1071 - thanks a lot for this useful
package)as an indication of possible self-thinning in an experimental tree
stand.

Problem: 
Results from the ks.test and the shapiro.test are not comparable (see example of
both tests). Tree height is a nice continuous variable. Sample size is around n=
250-350. Does anybode know about a problem in ks tests using a large sample size
and working with subsets (e.g file[,])? Comparing tests with qqplots, I would
appreciate shapiro, but I am wondering about the results from ks (test ist not
very sensitive, D=1, p=2.2e-16 many times?

Thanks
Sibylle

> shapiro.test(Biotree[Ac,]$Height2008)
	Shapiro-Wilk normality test

data:  Biotree[Ac, ]$Height2008 
W = 0.9908, p-value = 0.05175
> ks.test(Biotree[Ac,]$Height2008, "pnorm")
	One-sample Kolmogorov-Smirnov test

data:  Biotree[Ac, ]$Height2008 
D = 1, p-value < 2.2e-16
alternative hypothesis: two-sided 

Warning message:
In ks.test(Biotree[Ac, ]$Height2008, "pnorm") :
  cannot compute correct p-values with ties

--

In general Shapiro's normality test is more powerful than the KS. For this
specific case, I don't see the significantly different results from both
tests. The normality assumption in this example seems to be questionable.
-- 
View this message in context:
http://r.789695.n4.nabble.com/kstest-vs-shapirotest-tp2550192p2550243.html
Sent from the R help mailing list archive at Nabble.com.

The way that you called ks.test below your null hypothesis is that your data
comes from a normal distribution with mean 0 and standard deviation 1.  Now I am
not familiar with your data, but I would be very surprised in general to find a
population of trees where half of them had negative heights so it is not
surprising that the ks test shows your data to not be consistent with the null
of normal with mean 0, sd 1.

If you are interested in normality of populations based on samples greater than
250 then you should look into SnowsPenultimateNormalityTest() in the
TeachingDemos package, note however that the documentation is likely to be more
useful than the test itself.

-- 
Gregory (Greg) L. Snow Ph.D.
Statistical Data Center
Intermountain Healthcare
greg.snow at imail.org
801.408.8111

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of "Sibylle St?ckli"
> Sent: Wednesday, September 22, 2010 6:32 AM
> To: r-help at r-project.org
> Subject: [R] kstest vs shapirotest
> 
> Dear R-users
> 
> Idea:
> Analysing tree height frequency with hist(), normal distribution
> (ks.test & shapiro.test) and skewness (package e1071 - thanks a lot for
> this useful package)as an indication of possible self-thinning in an
> experimental tree stand.
> 
> Problem:
> Results from the ks.test and the shapiro.test are not comparable (see
> example of both tests). Tree height is a nice continuous variable.
> Sample size is around n= 250-350. Does anybode know about a problem in
> ks tests using a large sample size and working with subsets (e.g
> file[,])? Comparing tests with qqplots, I would appreciate shapiro, but
> I am wondering about the results from ks (test ist not very sensitive,
> D=1, p=2.2e-16 many times?
> 
> Thanks
> Sibylle
> 
> 
> > shapiro.test(Biotree[Ac,]$Height2008)
> 
> 	Shapiro-Wilk normality test
> 
> data:  Biotree[Ac, ]$Height2008
> W = 0.9908, p-value = 0.05175
> 
> > ks.test(Biotree[Ac,]$Height2008, "pnorm")
> 
> 	One-sample Kolmogorov-Smirnov test
> 
> data:  Biotree[Ac, ]$Height2008
> D = 1, p-value < 2.2e-16
> alternative hypothesis: two-sided
> 
> Warning message:
> In ks.test(Biotree[Ac, ]$Height2008, "pnorm") :
>   cannot compute correct p-values with ties
> 
> --
> 
> ______________________________________________
> 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.