[R-sig-eco] Clustering large data

[ONKELINX, Thierry]

Dear all,

Thanks for your responses. The biggest problem seems to be cast() for
the reshape package which could not handle the dataset. Peter's solution
using the mefa package worked fine. I found an other solution: table()
which works fine to crosstabulate presence-only data.

After crosstabulation I tried a few clusering methods. Agnes(), diana()
and hclust() gave a solution. Daisy() gave an out-of-memory error.

A follow-up question: I'm looking at the group membership with cutree().
It gives me something like:

> unique(cutree(test, k = 2:8))
     2 3 4 5 6 7 8
[1,] 1 1 1 1 1 1 1
[2,] 1 1 1 2 2 2 2
[3,] 1 1 2 3 3 3 3
[4,] 1 1 2 3 3 3 8
[5,] 1 1 2 3 3 4 4
[6,] 2 2 3 4 4 5 5
[7,] 2 2 3 4 5 6 6
[8,] 2 3 4 5 6 7 7

But I'm looking for a binary or dendrogram like coding of the group
membership. That would be more convenient for mapping the group
membership.

[1,] 111
[2,] 110
[3,] 1011
[4,] 1010
[5,] 100
[6,] 011
[7,] 010
[8,] 00

Any suggestions on that?

Thierry

------------------------------------------------------------------------
----
ir. Thierry Onkelinx
Instituut voor natuur- en bosonderzoek / Research Institute for Nature
and Forest
Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
methodology and quality assurance
Gaverstraat 4
9500 Geraardsbergen
Belgium 
tel. + 32 54/436 185
Thierry.Onkelinx at inbo.be 
www.inbo.be 

To call in the statistician after the experiment is done may be no more
than asking him to perform a post-mortem examination: he may be able to
say what the experiment died of.
~ Sir Ronald Aylmer Fisher

The plural of anecdote is not data.
~ Roger Brinner

The combination of some data and an aching desire for an answer does not
ensure that a reasonable answer can be extracted from a given body of
data.
~ John Tukey

-----Oorspronkelijk bericht-----
Van: r-sig-ecology-bounces at r-project.org
[mailto:r-sig-ecology-bounces at r-project.org] Namens Peter Solymos
Verzonden: dinsdag 7 oktober 2008 15:51
Aan: r-sig-ecology at r-project.org
Onderwerp: Re: [R-sig-eco] Clustering large data

Dear Thierry,

the 'mefa' package should do this, and I am also interested in the
testing of the package for such a large number of species. I have used
it before with 75K records, but only with ~160 species and 1052 sites.
So please let me know if it worked!

You can do the clustering like this (SAMPLES and SPECIES are the two
column in the long format, have to be the same length):

x <- mefa(stcs(data.frame(SAMPLES,SPECIES)))
cl <- hclust(dist(x$xtab))

Hope this works,

Peter

Peter Solymos, PhD
Department of Mathematical and Statistical Sciences
University of Alberta
Edmonton, Alberta, T6G 2G1
CANADA



On Tue, Oct 7, 2008 at 4:12 AM, ONKELINX, Thierry
<Thierry.ONKELINX at inbo.be> wrote:
> Dear all,
>
> We have a problem with a large dataset that we want to cluster. The
> dataset is in a long format: 1154024 rows with presence data. Each row
> has the name of the species and the location. We have 1381 species and
> 6354 locations.
> The main problem is that we need the data in wide format (one row for
> each location, one column for each species) for the clustering
> algorithms. But the 6354 x 1381 dataframe is too big to fit into the
> memory. At least when we use cast from the reshape package to convert
> the dataframe from a long to a wide format.
>
> Are there any clustering tools available that can work with the data
in
> a long format or with sparse matrices (only 13% of the matrix is
> non-zero)? If the work with sparse matrices: how to convert our
dataset
> to a sparse matrix? Other suggestions are welcome.
>
> We are working with R 2.7.2 on WinXP with 2 GB RAM. --max-mem-size is
> set to 2047M.
>
> Thanks,
>
> Thierry
>
>
>
------------------------------------------------------------------------
> ----
> ir. Thierry Onkelinx
> Instituut voor natuur- en bosonderzoek / Research Institute for Nature
> and Forest
> Cel biometrie, methodologie en kwaliteitszorg / Section biometrics,
> methodology and quality assurance
> Gaverstraat 4
> 9500 Geraardsbergen
> Belgium
> tel. + 32 54/436 185
> Thierry.Onkelinx at inbo.be
> www.inbo.be
>
> To call in the statistician after the experiment is done may be no
more
> than asking him to perform a post-mortem examination: he may be able
to
> say what the experiment died of.
> ~ Sir Ronald Aylmer Fisher
>
> The plural of anecdote is not data.
> ~ Roger Brinner
>
> The combination of some data and an aching desire for an answer does
not
> ensure that a reasonable answer can be extracted from a given body of
> data.
> ~ John Tukey
>
> Dit bericht en eventuele bijlagen geven enkel de visie van de
schrijver weer
> en binden het INBO onder geen enkel beding, zolang dit bericht niet
bevestigd is
> door een geldig ondertekend document. The views expressed in  this
message
> and any annex are purely those of the writer and may not be regarded
as stating
> an official position of INBO, as long as the message is not confirmed
by a duly
> signed document.
>
> _______________________________________________
> R-sig-ecology mailing list
> R-sig-ecology at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
>

_______________________________________________
R-sig-ecology mailing list
R-sig-ecology at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology

Dit bericht en eventuele bijlagen geven enkel de visie van de schrijver weer 
en binden het INBO onder geen enkel beding, zolang dit bericht niet bevestigd is
door een geldig ondertekend document. The views expressed in  this message 
and any annex are purely those of the writer and may not be regarded as stating 
an official position of INBO, as long as the message is not confirmed by a duly 
signed document.