- fixed bugs that were causing some examples to fail (due to an update in package loo)
- updated documentation of dwnorm, dwnorm2, fit_angmix and fit_incremental_angmix
- added an optional argument 'force_approx_const' to dvmcos
- added a function 'bestcriterion' to extract the value of best model selection criterion from stepwise fits.
- densityplot now uses lattice::wireframe instead of persp
- added alpha shading argument in contour.angmcmc.
- removed press.enter argument from plot.angmcmc, paramtrace and lpdtrace.
- changed BESSI, BESSI0, BESSI0_C and BESSI1. They're now all wrappers for R::bessel_i.
- changed "n"-cluster to "n"-component in print.angmcmc.
- Deleted AIC.angmcmc, BIC.angmcmc -- they can be directly computed using stats::AIC and stats::BIC.
- Combined all fitting functions with known component sizes into one single funciton called 'fit_angmix'. The previous model specific fitting functions e.g. 'fit_vmsinmix' etc. are retained for backward compatibility, but they all now wrap to the same 'fit_angmix' function. The 'fit_angmix' function includes a number of new features, such as multiple chains, an improved auto-tuning algorithm, permutation sampling, and option to impose restrictions on the component specific covariance parameters (such as fitting product mixtures).
- fit_stepwise_univariate and fit_stepwise_bivariate are now merged into fit_incremental_angmix.
- coda::as.mcmc.list method for mcmc objects is added.
- Added 'add_burnin_thin' as a post-processing function.
- Added 'bridge_sampler.angmcmc' method for bridgesampling::bridge_sampler
- WAIC computation is now done through the S3 method waic.angmcmc from package loo.
- Added loo.angmcmc function.
- fix_label now supports all methods for label.switching::label.switching
- A plot method for angmcmc objects is addedd.
- A naive rejection sampling method for vmsin and vmcos generation is added.
- Added two functions, one for evaluation of sample circular correlations for paired circular data, and one for evaluating population circular correlations and variances for vmsin, vmcos and wnorm2 models.
- Corrected a few typos in documentation.
- Corrected valgrind memory errors.
- Added arXiv link to vignette in DESCRIPTION.
- Added native routine registration.
- Corrected a few typos in documentation.
- Added option to choose hyper parameters in the Dirichlet prior for mixing proportion.
- Added a function to create logLik objects from angmcmc objects.
- Added a function to extract the best fitted model object from step-wise fits.
- Corrected bugs that caused build to fail with clang, MacOS, and Solaris.