The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.

Rtwalk 2.2.0

The algorithm is unchanged: it still follows Christen & Fox (2010) exactly, with the same default settings. Because random numbers are now drawn in a different order, chains differ from those of version 2.1.0 for the same seed.

Performance and memory

• The t-walk kernel is now implemented in C++ (via Rcpp). The overhead per iteration fell from about 65 to about 2 microseconds, so cheap targets sample about 30 times faster.
• Errors raised by log_posterior are caught without wrapping every evaluation in tryCatch(), which alone cost several times the evaluation of a simple target.
• The pure-R implementation is kept as a reference that follows the paper literally (twalk_control(engine = "R")); tests check that both engines produce the same chains draw by draw.
• New thin argument to store only every thin-th iteration.
• twalk_control(store_companion = FALSE) drops the companion trajectory, halving the memory used by the output.
• all_samples is no longer stored, which used as much memory as the two trajectories together. It is deprecated: accessing it rebuilds it, with a warning.

New features

• twalk_control() exposes the constants of the algorithm (n1, aw, at and the move probabilities), with the defaults of Christen & Fox (2010, Section 2.5), for sensitivity studies.
• twalk_continue() extends a previous run from the final state of the sampler.
• twalk_step() performs a single t-walk transition, so the kernel can be combined with Gibbs updates as described in Christen & Fox (2010, Section 5).
• The log posterior at the stored points is returned in log_posterior and companion_log_posterior.
• Parameter names: if x0 has names, log_posterior receives named vectors and the output columns are named.
• x0 and xp0 may be matrices with one row per chain, so that multiple chains start from dispersed points. A note is shown when all chains share the same starting points.
• Each chain now uses its own random-number stream, so multi-chain results are reproducible with set.seed() and do not depend on n_cores. With n_cores = 1, chains run sequentially without starting a cluster.
• Interrupting a sequential run returns the iterations completed so far.
• Errors raised by log_posterior at proposals are still treated as rejections, but a warning now reports how many occurred and the first error message.
• plot() method showing traces of the log posterior and of the parameters, one colour per chain.
• posterior::as_draws(), as_draws_array() and as_draws_df() methods.

Diagnostics

• calculate_diagnostics() and summary() now use the rank-normalized split-R-hat and the bulk and tail effective sample sizes of Vehtari et al. (2021); the estimates agree with those of the posterior package. R-hat is now also available for a single chain.
• summary() returns a summary.twalk data frame, printed with the sampler performance, instead of printing and returning the table invisibly.
• Diagnostics use the column names of the samples, if any.

Documentation

• New vignette reproducing the experiments of Christen & Fox (2010).
• New vignette comparing the t-walk with other general-purpose samplers.
• The validation vignette now shows its results next to the true values.

Other changes in this version

• Refactored sequential and parallel execution to use one shared chain engine.
• Limited progress-bar refreshes to approximately 100 per chain.
• Added an explicit error for identical initial points, which otherwise create a degenerate t-walk state.
• Added coda::as.mcmc() and coda::as.mcmc.list() methods for single and parallel t-walk results, with optional burn-in and companion conversion.
• Parallel workers load the same version of the package as the main process, including development versions loaded with pkgload::load_all().
• Added per-kernel proposal, acceptance, and actual-movement statistics for the Traverse, Walk, Blow, and Hop kernels.
• Corrected the Hop proposal-density helper so that only selected coordinates contribute to its continuous density.
• Added regression tests for Hop coordinate selection and the scalar affine invariance property established by Christen & Fox (2010).

Rtwalk 2.1.0

Algorithm correctness

• Corrected the Blow-kernel proposal density to include only coordinates selected for movement, matching Christen & Fox (2010).
• Bounded all Metropolis–Hastings acceptance probabilities to [0, 1] and moved their calculation to the log scale for numerical stability.
• Preserved accepted identity proposals from the original t-walk transition kernel while reporting them separately from actual state movements.
• Avoided repeated evaluations of the log posterior, reducing sequential execution to one evaluation per proposal plus the two initial evaluations.

Sampling and parallel execution

• Added reproducible independent random-number streams for parallel chains; set.seed() now controls both sequential and parallel runs.
• Parallel chains now preserve the supplied initial points and support constrained parameter spaces correctly.
• Parallel execution now exports direct and recursively referenced log-posterior dependencies to worker processes.
• Added consistent n_chains and total_iterations metadata in sequential and parallel results.

Output and diagnostics

• Added separate samples and companion_samples trajectories. The legacy all_samples field remains available for compatibility but is not treated as one time-ordered chain.
• Added move_rate and no_move_rate alongside the Metropolis–Hastings acceptance_rate.
• Applied burn-in independently to each parallel chain and computed effective sample sizes without joining chain boundaries.
• Unified sample and burn-in validation across diagnostics and visualization.

Interface and validation

• Added the public show_progress argument. Additional arguments, including an argument named internal_call, are now passed exclusively to log_posterior.
• Added clear validation for iteration counts, chains, cores, initial points, log-posterior return values, burn-in fractions, and parameter names.
• Expanded regression tests for kernel mathematics, parallel reproducibility, constrained supports, dependency export, diagnostics, and edge cases.

Rtwalk 2.0.2

• Added a comprehensive unit test suite using testthat, achieving 100% test coverage.
• Added testthat and covr to Suggested dependencies.

Rtwalk 2.0.1

• Added S3 methods for print() and summary() for t-walk output objects.
• Improved diagnostics reporting and user-facing summaries.
• Minor internal refactoring with no changes to core algorithm behavior.

Rtwalk 2.0.0

• Reintroduced the t-walk MCMC algorithm to CRAN.
• Modernized the codebase to comply with current CRAN policies.
• Updated documentation and examples.
• Ensured compatibility with recent versions of R.

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.