R help - Jul 2008 - p-value for Nonmetric Multidimentional Scaling?

Dear R-helpers,

I am running metaMDS in the vegan package, which uses isoMDS in MASS, to perform
Nonmetric Multidimentional Scaling (NMDS).

I have seen some authors report a p-value for the NMDS ordination based on
randomization of the dataset. As I understand it this is meant to compare the
stress in your dataset to multiple runs of randomized data.
I do not see a way to perform such a test in vegan or MASS.

So my questions are:
Is this necessary? and Does R have a function to do this?

Thanks in advance for your help,
Michael

You need to ask yourself a number of questions, e.g.

What is the hypothesis you wish to test?
What is the test statistic you wish to use to test it?
How can I get some information on where my value of that statistic sits
with respect to its null hypothesis distribution?

p-values do not exist independently of hypothesis tests.


(It would also be a coutresy to the list to give your name and
affiliation, as the posting guide suggests.) 


Bill Venables
CSIRO Laboratories
PO Box 120, Cleveland, 4163
AUSTRALIA
Office Phone (email preferred): +61 7 3826 7251
Fax (if absolutely necessary):  +61 7 3826 7304
Mobile:                         +61 4 8819 4402
Home Phone:                     +61 7 3286 7700
mailto:Bill.Venables at csiro.au
http://www.cmis.csiro.au/bill.venables/ 

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
On Behalf Of Michael Denslow
Sent: Saturday, 5 July 2008 2:14 PM
To: r-help at r-project.org
Subject: [R] p-value for Nonmetric Multidimentional Scaling?

Dear R-helpers,

I am running metaMDS in the vegan package, which uses isoMDS in MASS, to
perform Nonmetric Multidimentional Scaling (NMDS).

I have seen some authors report a p-value for the NMDS ordination based
on randomization of the dataset. As I understand it this is meant to
compare the stress in your dataset to multiple runs of randomized data.
I do not see a way to perform such a test in vegan or MASS.

So my questions are:
Is this necessary? and Does R have a function to do this?

Thanks in advance for your help,
Michael

______________________________________________
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.

On Fri, 2008-07-04 at 21:14 -0700, Michael Denslow
wrote:
> Dear R-helpers,
> 
> I am running metaMDS in the vegan package, which uses isoMDS in MASS,
> to perform Nonmetric Multidimentional Scaling (NMDS).
> 
> I have seen some authors report a p-value for the NMDS ordination
> based on randomization of the dataset. As I understand it this is
> meant to compare the stress in your dataset to multiple runs of
> randomized data.
> I do not see a way to perform such a test in vegan or MASS.
> 
> So my questions are:
> Is this necessary? and Does R have a function to do this?
No, but the following discussion on the r-forge repository for vegan
presents some ways that an analysis of this type could be done using
metaMDS:

http://r-forge.r-project.org/forum/forum.php?thread_id=586&forum_id=194

But read Jari's comments carefully. If your data are species, then
consider whether permuting row, columns or both is appropriate when they
don't preserve the row or column totals. So a question to ask yourself
is, is permuting rows, columns or both going to derive the appropriate
Null hypothesis for the test you have in mind?

Please feel free to follow this up in the forum post I refer to.

All the best,

G
> 
> Thanks in advance for your help,
> Michael
> 
> ______________________________________________
> 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.