[Fwd: [R-sig-Geo] spdep 0-4.13]

[Paolo Cavallini]

FYI

-------- Messaggio Originale  --------
Oggetto: [R-sig-Geo] spdep 0-4.13
Data: Thu, 3 Jan 2008 12:56:06 +0100 (CET)
Da: Roger Bivand <Roger.Bivand@...>
Rispondi-a: Roger.Bivand@...
A: r-sig-geo@...

A new release of spdep has been published on CRAN and will spread to a
mirror near you soon. It includes faster sparse matrix computation in
maximum likelihood model fitting, and code for exact global and local
Moran's I tests contributed by Markus Reder.

Please note two issues: spdep now depends on the spam package (another
sparse matrix implementation), so that just updating spdep will not be
enough, and spam needs to be installed too, unless you already have it
(users of the fields package will probably have it if you update
regularly).

The second issue is that the ade4 package is currently unable to load
spdep in functions that need this dependence, because of spdep's new
dependence on spam. We are working on this, and a suggested fix for spam
has been sent to its maintainer. So ade4 users who do not need extra spdep
functionality may choose not to update spdep until this issue has been
resolved.

Please let me know of any difficulties,

Happy New Year!

Roger

--
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: [phone]; fax [phone]
e-mail: Roger.Bivand@...

_______________________________________________
R-sig-Geo mailing list
R-sig-Geo@...
https://stat.ethz.ch/mailman/listinfo/r-sig-geo


--
Paolo Cavallini, see: http://www.faunalia.it/pc

_______________________________________________
AniMov mailing list
AniMov@...
http://www.faunalia.com/cgi-bin/mailman/listinfo/animov

[Mathias Tobler]

I am analyzing some data from GPS collars using the Brownian Bridge
Movement Model in adehabitat and was wondering what exactly sig1
represents. Using the maximum likelihood method I usually get values
in the range from 1 to 4. Reading the documentation it says: "The
parameter sig1 is related to the speed of the animal, and describes
how far from the line joining two successive relocations the animal
can go during one time unit (here the time is measured in second)".
Taking a value of 3 m in one second for sig1 that would mean my
animal moves at 10 km/h, definitely too fast. For the puechcirc the
value is 6.32 which translates to 22 km/h. Reading the Horne et. al.
2007 paper they give values for sig1 of several 100 m. Is sig1
implemented differently in adehabitat?

Thanks,

Mathias



_______________________________________________
AniMov mailing list
AniMov@...
http://www.faunalia.com/cgi-bin/mailman/listinfo/animov

[Clément Calenge]

Dear Mathias,

Sorry for this late reply, but I was not at my office during the two
last weeks, and could not reply to your email. I hope this is not too
late for you.

> I am analyzing some data from GPS collars using the Brownian Bridge
> Movement Model in adehabitat and was wondering what exactly sig1
> represents. Using the maximum likelihood method I usually get values
> in the range from 1 to 4. Reading the documentation it says: "The
> parameter sig1 is related to the speed of the animal, and describes
> how far from the line joining two successive relocations the animal
> can go during one time unit (here the time is measured in second)".
> Taking a value of 3 m in one second for sig1 that would mean my
> animal moves at 10 km/h, definitely too fast. For the puechcirc the
> value is 6.32 which translates to 22 km/h. Reading the Horne et. al.
> 2007 paper they give values for sig1 of several 100 m. Is sig1
> implemented differently in adehabitat?

First, note that the functions liker and kernelbb (adehabitat) use sig1,
while Horne et al. use sig1^2. Then, note the key word in the help page
of the function kernelbb, concerning the parameter sig1:

"*related* to the speed of the animals; it can be estimated by the
function 'liker'"

Thus, sig1 is not the speed of the animal, but is *related* to the speed
of the animal. The reason why this parameter is ambiguous is that its
interpretation is rather complicated. An important aspect is that Horne
et al. call sig1^2 the "Brownian motion variance parameter". It is even
not the variance of the position, but a parameter used to compute the
variance (which is itself related to the speed). The variance of the
position is S² = (sig1^2) * t*(T-t) / T, where T is the duration of the
trajectory between the two relocations and t the time at the current
position.

S is the standard deviation of the position, and is therefore measured
in meters (if the original data were measured in meters), S^2 is
therefore measured in squared meters. On the other hand, t is measured
in seconds, and (T-t)/t has no dimension (seconds divided by seconds).
Consequently, the units of sig1^2 are therefore m²/seconds, i.e. the
units of sig1 are m/sqrt(seconds). The value of sig1^2 in the paper of
Horne et al (at least concerning the black bear) is in squared meters
per minute (explaining the difference in magnitude).

Therefore your results are sig1 = 6.32 m/sqrt(seconds). As you can see,
this parameter is *not* a speed! AFAIK, given the present state of
research on Brownian bridge kernel, it is difficult to find a biological
interpretation to this parameter (which is different from sig2, for
which the interpretation is clear: it is the imprecision on the
relocations, measured in meters or any other units found in the original
data). The only thing that can be said is that "it is related to the
speed of the animal". This relationship is S² = (sig1^2) * t*(T-t) / T,
S² being related to the speed of the animal.

But you may be right on one aspect: this is not very clear on the help
page of the function. For the next version of adehabitat, I will try to
make it clear that sig1 is not a speed, but is related to the speed. I
will give some of the above information on the help page.
Hope this helps,


Clément.

--
Clément CALENGE
LBBE - UMR CNRS 5558 - Université Claude Bernard Lyon 1 - FRANCE
tel. [phone]
fax. [phone]


_______________________________________________
AniMov mailing list
AniMov@...
http://www.faunalia.com/cgi-bin/mailman/listinfo/animov