- This is a major update, with new functions for estimating the Bayesian Mallows model using sequential Monte Carlo. The methods are described in the vignette titled "SMC-Mallows Tutorial".
- Removed a large number of dependencies by converting to base R code. This will make the package easier to install across a range of systems, and less vulnerable to changes in other packages.
- Incorporates changes since 1.0.3, and also remove PLMIX from Imports.
- Fixed bug which caused plot_top_k to fail when plotting clusters.
- Improved the default value of rel_widths argument to plot_top_k.
- Wrote unit tests to check that the bugs don't appear again.
- Fixed bug which caused importance sampling to fail when running in parallel.
- Fixed issue with error message when trying to plot error probability when compute_mallows has not been set up to compute error probability.
- Increased number of unit tests.
- Fixed critical bug which caused results to be wrong with more than one mixture component in compute_mallows() and compute_mallows_mixtures(). Thanks to Anja Stein for discovering the bug.
- Function generate_initial_ranking() now has two additional options for generating random initial rankings. This can help with convergence problems, by allowing the MCMC algorithm to run from a range of different starting points.
- Fixes a bug in lik_db_mix and expected_dist, in which the scaling parameter used a different parametrization than the rest of the package. All functions in the package now use consistent parametrization of the Mallows model, as stated in the vignette.
- Function for computing likelihood added.
- Options for dealing with missing values added, and documentation now states how missing values are dealt with.
- Function rank_freq_distr added, for computing the frequency distribution of ranking patterns.
- Function rank_distance added, for computing the distance between rankings.
- Function expected_dist added for computing expectation of several metrics under the Mallows model.
- Function compute_consensus now includes an option for computing consensus of augmented ranks.
- Fixed bug in predict_top_k and plot_top_k when using aug_thinning > 1.
- Updated README and vignette.
- Updating a unit test to make sure BayesMallows is compatible with dplyr version 1.0.0.
- Improvement of plotting functions, as noted below.
- plot.BayesMallows and plot_elbow no longer print titles automatically.
- assess_convergence no longer prints legends for clusters, as the cluster number is essentially arbitrary.
- Added CITATION.
- Updated test of random number seed.
- Implements all fixes since version 0.3.1 below.
- Fixed typo on y-axis label of elbow plot.
- Fixed an issue which caused the cluster probabilities to differ across platforms, despite using the same seed. https://stackoverflow.com/questions/54822702
- Fixed a bug which caused
compute_mallowsnot to work (without giving any errors) whenrankingscontained missing values. - Fixed a bug which caused
compute_mallowsto fail whenpreferenceshad integer columns.
- Changed the name of
save_individual_cluster_probstosave_ind_clus, to save typing.
- Added a user prompt asking if the user really wants to save csv files, when
save_individual_cluster_probs = TRUEin compute_mallows. - Added
alpha_max, the truncation of the exponential prior foralpha, as a user option incompute_mallows.
- Added functionality for checking label switching. See
?label_switchingfor more info.
- The internal function
compute_importance_sampling_estimatehas been updated to avoid numerical overflow. Previously, importance sampling failed at below 200 items. Now it works way above 10,000 items.
- This is an update of some parts of the C++ code, to avoid failing the sanitizer checks clang-UBSAN and gcc-UBSAN.
- See all bullet points below, since 0.2.0.
generate_transitive_closure,generate_initial_ranking, andgenerate_constraintsnow are able to run in parallel.- Large changes to the underlying code base which should make it more maintainable but not affect the user.
estimate_partition_functionnow has an option to run in parallel, leading to significant speed-up.
- Implemented the Bernoulli error model. Set
error_model = "bernoulli"incompute_mallowsin order to use it. Examples will come later.
- Added parallelization option to
compute_mallows_mixturesand addedparallelto Suggests field.
- Deprecated functions
compute_cp_consensusandcompute_map_consensushave been removed. Usecompute_consensusinstead.
- Clusters are now
factorvariables sorted according to the cluster number. Hence, in plot legends, "Cluster 10" comes after "Cluster 9", rather than after "Cluster 1" which it used to do until now, because it was acharacter. plot.BayesMallowsno longer contains print statements which forces display of plots. Instead plots are returned from the function. Usingp <- plot(fit)hence does no longer display a plot, whereas usingplot(fit)without assigning it to an object, displays a plot. Until now the plot was always shown forrhoandalpha.
compute_mallowsandsample_mallowsnow support Ulam distance, with argumentmetric = "ulam".- Slimmed down the vignette significantly, in order to avoid clang-UBSAN error caused by running the vignette (which was then again caused by
Rcpp, cf. this issue). The long vignette is no longer needed in any case, since all the functions are well documented with executable examples.
- New release on CRAN, which contains all the updates in 0.1.1, described below.
Rankclusterpackage has been removed from dependencies.
- Fixed bug with Cayley distance. For this distance, the computational shortcut on p. 8 of Vitelli et al. (2018), JMLR, does not work. However, it was still used. Now, Cayley distance is always computed with complete rank vectors.
- Fixed bug in the default argument
leap_sizetocompute_mallows. It used to befloor(n_items / 5), which evaluates to zero whenn_items <= 4. Updated it tomax(1L, floor(n_items / 5)). - Added Hamming distance (
metric = "hamming") as an option tocompute_mallowsandsample_mallows.
- Updated
generate_initial_ranking,generate_transitive_closure, andsample_mallowsto avoid errors when packagetibbleversion 2.0.0 is released. This update is purely internal.
- Objects of class
BayesMallowsandBayesMallowsMixturesnow have default print functions, hence avoiding excessive amounts of informations printed to the console if the user happens to write the name of such an object and press Return. compute_mallows_mixturesno longer setsinclude_wcd = TRUEby default. The user can choose this argument.compute_mallowshas a new argumentsave_clus, which can be set toFALSEfor not saving cluster assignments.
assess_convergencenow automatically plots mixtures.compute_mallows_mixturesnow returns an object of classBayesMallowsMixtures.
assess_convergencenow adds prefix Assessor to plots whenparameter = "Rtilde".predict_top_kis now an exported function. Previously it was internal.
compute_posterior_intervalsnow has defaultparameter = "alpha". Until now, this argument has had no default.- Argument
typetoplot.BayesMallowsandassess_convergencehas been renamed toparameter, to be more consistent.
- Argument
save_augment_datatocompute_mallowshas been renamed tosave_aug. compute_mallowsfills in implied ranks when an assessor has only one missing rank. This avoids unnecessary augmentation in MCMC.generate_rankingandgenerate_orderingnow work with missing ranks.
Argument cluster_assignment_thinning to compute_mallows has been renamed to clus_thin.
Change the interface for computing consensus ranking. Now, CP and MAP consensus are both computed with the compute_consensus function, with argument type equal to either "CP" or "MAP".