This page is no longer up-to-date, visit datacloning.org instead
Welcome to dcr - Data Cloning in R project!
Data cloning (DC) uses Bayesian MCMC to make maximum likelihood inference of complex hierarchical models. The bundle includes basic infrastructure for DC with parallel computing support, and more specialized packages for ecology.
The project summary page you can find here.
The stable dclone release you can find at CRAN.
Documentation
- Short introduction: Péter Sólymos, 2010. dclone: Data Cloning in R.
The R Journal, 2(2):29-37, December 2010. URL
http://journal.R-project.org/
[bib]
[pdf]
[local copy]
[R code] (Official citation)
- Reference manual: [pdf] (Detailed technical documentation of the package from CRAN)
- Additional resources: tutorials (under development),
worked examples (also available via the dcmle package), extras
(not fully supported and experimental features).
Infrastructure for dclone
Because dclone is an R package, installing R is absolutely necessary.
The choice of the preferred BUGS program is up to the user.
- R (>= 2.14.0): download from
CRAN.
- WinBUGS (>= 1.4): download
from here, don't forget
the patch and the immortality key, the R2WinBUGS and
coda R packages are required for using WinBUGS from within R.
- OpenBUGS: download
from here, it requires the
BRugs (>= 0.3-2) (or the R2OpenBUGS) R package .
- JAGS (>= 3.0.0): download
from here,
the rjags and coda
R packages are required for using JAGS from within R.
The dclone package itself depends on the coda (>= 0.13), R2WinBUGS, and parallel
which are all available from CRAN.
The dclone package also suggests to have the rjags (>= 3-2), snow,
rlecuyer, rsprng, BRugs R packages (the latter two might not be available for all platforms).
The rjags package is suggested for using JAGS.
The rjags dependency
of dclone was removed so that other functionality of the package
can be used without JAGS being installed (rjags can't load without JAGS).
The easiest way to install the stable release and all required packages
at once is to type this after opening R:
install.packages(c("dclone", "rjags"))
The development version can be installed as:
install.packages("dclone", repos = "http://r-forge.r-project.org")
Parallel computing
Parallel computing is supported via clusters or multiple cores.
Clusters defined by the snow or parallel
packages can be used to run multiple parallel MCMC chains by JAGS.
Forking type parallelism via the parallel package
is also supported.
Parallel computing for data cloning is provided either via
parallel chains, `size balancing'
(see this tutorial
for a general overview), or a combination of the the two.
Size balancing is available for WinBUGS/OpenBUGS as well.
Specialized packages in this repository
- dcmle: Hierarchical Models Made Easy with Data Cloning.
S4 classes around infrastructure provided by the dclone package to make package development
with data cloning for hierarchical models easy as a breeze.
[get it from CRAN]
- pbapply: a lightweight package that adds
progress bar to vectorized R functions ('*apply'). The implenentation can easily be added
to functions, where showing the progress is useful for the user (e.g. bootstrap).
[get it from CRAN]
- sharx: data sets and SAR, SARX, HSAR and HSARX
models as described in Solymos and Lele (in press).
[get it from CRAN]
- ResourceSelection: Resource Selection (Probability) Functions
for use-availability wildlife data as described in Lele and Keim
(2006, Ecology 87, 3021–3028), and Lele (2009, J. Wildlife Management 73, 122–127).
[get it from CRAN]
- detect: Analyzing wildlife data with detection error.
The package implements models to analyze site occupancy and count data models
with detection error.
[get it from CRAN]
- PVAClone: Likelihood based population viability
analysis in the presence of observation error and missing data.
The package can be used to fit, compare, predict, and forecast
various growth model types using data cloning.
[get it from CRAN]