Skip to content

nested Kruskal Wallis

3 messages · Mike Dunbar, Erika Mudrak, Sarah Goslee

Original
Mike Dunbar
# 
Dear Jude

I'm trying to reconcile where you say "it doesn't meet the assumption of homogeneity of variances" with "sample size within fields is small and unequal". With small sample sizes you can easily generate dummy samples with widely differing variances even where the process generating the data has a single variance. It's difficult to say without more information, but I'd be tempted just to use a linear mixed model to analyse the data. 

regards

Mike D
Hi,

I'm trying to analyse a dataset on the size of beetles collected in
different crop types.  Crop type is the fixed effect, and field is a
random factor, nested in crop.  Although my data is normal after
transformation, it doesn't meet the assumption of homogeneity of
variances.  In addition, sample size within fields is small and
unequal (n=2-10).  So, I'm trying to figure out if I can run a nested
Kruskal Wallis type analysis, since this is more robust to both
problems than an ANOVA, even with permutation.  I've been told it
might be possible to run such an analysis within the coin package, but
I can't figure out how to do this -methods for blocked designs are
mentioned in the documentation, but not nested designs.

I'd appreciate any help or thoughts!

Cheers, Jude Phillips

_______________________________________________
R-sig-ecology mailing list
R-sig-ecology at r-project.org 
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
Erika Mudrak
# 
Hi everyone,

I am trying to investigate how well a certain (quirky) quadrat layout will uncover underlying spatial phenomenon of plant locatons.  I am using the geoR package to generate variograms, but I am noticing some interesting trends that may be artifacts of the sampling design rather than the natural phenomenon.  

I would like to conduct simulations where I generate a known spatial pattern, superimpose the quadrat layout, sample from that, and analyze the resulting variogram.  I see that geoR has a sample.geodata() function, but I don't want to pick random samples, I want to easily determine what my data would look like if I could only see it through the quadrat layout.   Is there any function where I can easily do this?

[I realize that there is a lot of literature about preferred quadrat layouts, but the data was taken before I arrived at school, handed to me, and if I can figure out how to work with it, then I get a prize - a degree!]

Thanks
Erika

-------------------------------------------
Erika Mudrak
Graduate Student
Department of Botany
University of Wisconsin-Madison
430 Lincoln Dr
Madison WI, 53706
608-265-2191
mudrak at wisc.edu
Sarah Goslee
# 
Erika,

This article: http://dx.doi.org/10.1007/s11258-005-3495-x
 Behavior of Vegetation Sampling Methods in the Presence of Spatial
Autocorrelation
Plant Ecology Volume 187, Number 2 / December, 2006
used the gstat package to generate spatial patterns, and then sampled them using
different quadrat layouts.

The basic workflow:
Use gstat to generate fake vegetation patterns.
Sample the data using your quadrat layout.
Analyze as if it were real data.

You need to decide in advance how big a pixel is, so how many
pixels per quadrat your samples are: that will depend on the size
of the plants you are working on.

More detailed help would require knowing more about what you
are trying to do.

Sarah
On Thu, Nov 20, 2008 at 2:34 PM, Erika Mudrak <mudrak at wisc.edu> wrote: