This is a first release of the bimark package: which will be mostly made of
two parts:
- Superficial "shallow" analysis interface, allowing user to generate, manipulate and visualize bilateral mark-recapture data. Written in R.
- Underlying "deep" analysis interface, allowing user to perform a bayesian analyis of these data based on their underlying polytope structure. Written in R, JAGS and C++.
Althought everything has already been written and run during my past internship in 2015, this first release only ships the "shallow" part. This way, user may:
- assert that.. yes, I am currently working on the overall packaging on this code (mostly at nights and weekends).
- try out the first shallow functionnalities:
- get into the logic
- provide feedback
- provide bugreports
- provide feature requests
In a nutshell, this is an early stage of the project: visit the repo at http://github.com/iago-lito/bimark/.
Bimark is dedicated to analysis and simulation of Capture-Mark-Recapture data in a particular context where the left and right sides of captured individuals cannot always be matched with one another. We call this the "bilaterality" problem.
For instance, dolphins may be photo-identified with their dorsal fin. But depending on whether they exhibit deep marks on the edges or shallow marks on the side, one may or may not be able to tell whether a left- and a right-picture correspond to the same fin.
This situation has been formalized in detail by Link et al. 20101. It has been adapted to the problem of bilaterality both by McClintock et al. 20132 and Bonner et al. 20133.
The solution we have found to deal with it is mostly inspired from McClintock and Bonner. Our improved Bayesian sampling algorithm is described in detail in our M2 report, 20154, along with our notations.
Installing bimark, running the tests and building the documentation should be
as easy as:
> library(devtools)
> install_github("iago-lito/bimark", build_vignettes=TRUE)
.. but it is NOT XD
At least not yet. Sorry.
As a temporary solution, please perform this whole development procedure:
$ git clone https://github.com/iago-lito/bimark
Then in R:
> library(devtools)
> library(testthat)
> setwd("bimark")
> document()
> use_testthat()
> setwd("..")
> install("bimark", dependencies=TRUE, build_vignettes=TRUE)
> library(bimark)
> test_package("bimark")
> run_examples("bimark")
If everything runs fine, then you're done!
To generate a model based on simulated data, try:
> library(bimark)
> m <- BimarkSimulationModel(N=20, T=5)
The bimark model object is just a list. Access data with $:
> m$n # number of capture histories actually observed
> m$LR # number of observed right-histories
> m$iOmega # ids of all histories relevant to these data
> SeeHist(m$iOmega) # visualize histories and their ids
To generate a model based on actual observation data, use:
> myData <- example.M
> m <- BimarkObservationModel(myData)
> print(m) # unknown number of individuals, since sides haven't been matched
To retrieve matrices information from the model, feed dedicated methods with it:
> GetOmega.B(m) # all unobservable histories that may underlie these data
> get.A(m) # observation matrix sorted in polytope order
> get.B(m) # kernel of A matrix generating the polytope
Get information with:
> ?BimarkSimulationModel
> ?GetOmega.B
> ?get.A
If anything goes wrong (and things will go wrong), please file an
issue report on the repo.
We're also pleased to read your feature requests.
Iago-lito (https://github.com/iago-lito)
Dr. Olivier Gimenez, ([email protected])
This package is licensed under the GPL v3 license. © 2016 Bimark contributors
Footnotes
-
Link et al. 2010: doi:10.1111/j.1541-0420.2009.01244.x ↩
-
McClintock et al. 2013: doi:10.1890/12-1613.1 ↩
-
Bonner et al. 2013: doi:10.1111/biom.12045 ↩