LoginSign Up
← Back to all sparks
J

jSDM

INFRA · APIS
Velocity0.0

Joint Species Distribution Models

github.com/ghislainv/jsdm ↗

Joint species distribution models in Gibbs-sampled C++, quiet since 2023.

species-distribution-modelsbayesiangibbs-samplingecologyrcpp
◆Current state
jSDM fits joint species distribution models by Gibbs sampling, with the sampler written in C++ against GSL and Armadillo and exposed through binomial probit, binomial logit, Poisson log and Gaussian entry points. The 0.2 line extended it with species traits, constrained factor loadings and residual-association plots. The one entry since 2023 carries only a compare link.
◆Where it's heading
The package built out its model family quickly and then stopped: five of the six visible entries are stamped the same day as a backfilled archive, and the only later release says nothing about its contents. The direction the 0.2 line was heading, toward trait-mediated species effects and better convergence on latent variable models, has no visible continuation.
◆Prediction
The feed does not support a confident prediction; the latest entry publishes no notes, so whether the package is still developing or only being kept CRAN-clean cannot be read from it.

◆Recent moves

  1. 9mo ago

    jSDM CRAN release v0.2.7: Joint Species Distribution Models

    The only release since the 2023 archive, and its body is a compare link with no notes. Nothing about the direction of the package can be read from this entry.

    View source ↗
  2. 3y ago

    Species traits, constrained loadings and association plots

    The widest release of the 0.2 line: a hierarchical level for species-trait and environment interactions across the probit, logit and Poisson models, a constrained-species variant to help latent variable models converge, and plot_associations() for species-species associations. The block-suffixed function names are dropped in the process.

    View source ↗
  3. 3y ago

    First release: C++ Gibbs sampler for joint species models

    ⚡ SPARK

    The first version of the package, implementing the Warton et al. joint species distribution framework with a Gibbs sampler in C++ over GSL and Armadillo. Everything in the 0.2 line extends this base.

    View source ↗
  4. 3y ago

    Residual correlation functions can filter to significant values

    get_residual_cor() and plot_residual_cor() gain the option to consider only significant correlations, which matters once the species matrix is large enough that everything correlates with something.

    View source ↗
  5. 3y ago

    jSDM_gaussian() fits continuous data with overdispersion

    Adds jSDM_gaussian() for continuous responses with an overdispersion parameter, extending the package past presence-absence and count data, and moves the Madagascar vignette onto terra now that raster and sp are deprecated.

    View source ↗
  6. 3y ago

    Four versions of CRAN check corrections

    Covers v0.2.2 through v0.2.6 as minor corrections to pass CRAN checks, with no functional change across four version bumps.

    View source ↗