[AniMov] Distance sampling

Paolo Cavallini cavallini at faunalia.it
Wed Jun 1 07:25:37 CEST 2005


Hi all.
Somewhat related to AniMove: I thought of reproducing the functionality of 
Distance sampling software for density estimation 
(http://www.ruwpa.st-and.ac.uk/distance/) under R. This will have a number of 
advantages, including multiplatform availability, etc. The Distance developer 
was positive about this (see below). Is there anybody interested? As we use 
Distance very often, and we have many sample data set, we can easily do the 
testing, if someone is willing to play with the code.
All the best.
pc

----------  Forwarded message  ----------

Subject: Re: Distance and Linux
Date: 14:39, mercoledì 04 maggio 2005
From: "Distance Development" <distance at mcs.st-and.ac.uk>
To: "cavallini at faunalia.it" <cavallini at faunalia.it>

Hi Paolo,

Using the mrds r libary in Distance 5, you can analyze double and single
observer data - the main thing it does not do that the CDS and MCDS engines
allow is adding covarates to the detection function -- although that will
hopefully come at some point in the future.  The main problem with the mrds
package is that it currently has no R help -- again, hopefully that will
come in the future.  To learn how to use it at present your best option is
to run some mrds analyses in debug mode and check out the R code generated.

The CDS and MCDS analysis engine (mcds.exe) is compiled FORTRAN code.  I'd
be happy to send you the code and you could recompile it in FORTRAN -- but
it uses an external numerical library -- the IMSL library.  This is, I
understand, available for Linux (see supported computing platforms in
http://www.vni.com/products/imsl/fortran/overview.html) so you could always
get it.  Of course you'd need a FORTRAN complier too -- and I think the CDS
code uses some F90 code (to make variable sized arrays) so that might cause
problems -- not sure if it's standard or not (I used Compaq Visual Fortran
in Windows as the complier).  Might take a fair bit of work to get going.
Not too hard to link to this once it's compiled from within R, if you want
to (I have some code that does this in windows - probably similar code would
work in Linux).

I'm afraid I'm rather too busy to offer much by way of support to either of
these two ventures, although I'm happy to offer pointers where I can, and
would be really glad if you were willing to take it on.  Would be great if
you could write a brief guide to using Distance under Linux that we could
then put on the Distance web site for others to use.

Hope this helps some?

 - Len

On Mon, 2 May 2005 13:45:09 +0200, Paolo Cavallini wrote:
>Dear Sir,
>a few years ago we moved our computing infrastructure to Linux, with 
excellent
>results. We find somewhat ironical that the only software we need windows for
>is a non-commercial one, i.e. distance. Given that most of the analyses done
>by Distance are already implemented in R, I think the options to make us
>Linux users able to use Distance (even if without a dedicated GUI) would not
>be too difficult, and this would have several other advantages (e.g. the
>possibility to link it to a full GIS like GRASS).
>We would appreciate your comments on this, and would be glad of helping if
>necessary.
>All the best.

-- 
Paolo Cavallini
cavallini at faunalia.it   www.faunalia.it   www.faunalia.com
Piazza Garibaldi 5 - 56025 Pontedera (PI), Italy   Tel: (+39)348-3801953



More information about the AniMov mailing list