NEWS in package ‘mclust’

---

Changes in version 6.1.1

• Corrected computation of df for MclustDA and EDDA classification models.
• Bug fix on sim() when EVV, EVE, and VVE models for G = 1.
• Bug fix on summary.MclustBootstrap() when computing confidence intervals for G = 1.

Changes in version 6.1

• Added logsumexp() and softmax() functions as a wrapper to efficiently implementations written in Fortran code.
• Substituted R code in several parts of the package with the above Fortran-based functions to compute densities and posterior probabilities from Gaussian mixtures.
• Changes on me.weighted() to use convergence criterion as in other mclust functions and improved efficiency by using the above mentioned Fortran-based functions. This also brings computational improvements in the weighted likelihood bootstrap implemented in MclustBootstrap(..., type = "wlbs").
• Bug fix on MclustDA() when number of obs is less than number of vars.
• Bug fix on MclustBootstrap(object, ..., type = "pb") when object is of class densityMclust.

Changes in version 6.0.1

• Changed initialization in MclustSSC() for components of unlabeled data via k-means.
• Corrected output of summary.MclustSCC() for components of unlabeled data.
• Updated citation info with reference to book published by Chapman & Hall/CRC

Changes in version 6.0.0

• Major release of mclust accompanying the upcoming book by Chapman & Hall/CRC.

Changes in version 5.4.11

• Added summary.crimcoords() method and removed argument plot from crimcoords() function call.

Changes in version 5.4.10

• Updated banner on startup.
• Updated info on man page for datasets diabetes, wdbc, and thyroid.
• Std. error for cross-validation in cvMclustDA() uses formula for the weighted standard deviation with weights given by folds size.
• Fix .Rd files.

Changes in version 5.4.9

• Added crimcoords() to compute discriminant coordinates or crimcoords.
• Fixed man page for cvMclustDA().

Changes in version 5.4.8

• densityMclust() by default draw a graph of the density estimate.
• Fixed a bug in computing mixture density if the noise component is present.
• Changed default behaviour of hc() when called to perform agglomerative hierarchical clustering instead of using for EM initialization.
• The default mclust.options("hcModelName") now returns only the model to be used.
• Changed default partition argument of hc() function by adding dupPartion() to remove data duplicates.
• Added checks to mclustBootstrapLRT() to stop if an invalid modelName is provided or a one-component mixture model is provided.
• Extended the functionality of cvMclustDA() by including as cross-validated metrics both the classification error and the Brier score.
• Updated info on dataset man pages.

Changes in version 5.4.7

• Updated plot method (dendrogram) for hierarchical clustering --- now based on classification likelihood.
• Added MclustSSC() function (and related print, summary, plot, and predict, methods) for semi-supervised classification.
• Exchanged order of models VEE and EVE to account for increasing complexity of EVE.
• Added cex argument to clPairs() to control character expansion used in plotting symbols.
• em() and me() have now data as first argument.

Changes in version 5.4.6

• Fixed issues with source Fortran code with gfortran 10 as reported by CRAN.
• Clean code of hcCriterion().
• Replaced CEX argument in functions with standard base graph cex argument.
• Removed ylim argument in function; it can be passed via ....
• MclustDA models use the default SVD transformation of the data for initialization of the EM algorithm.
• Added icl criterion to object returned by Mclust().
• Fixed number of pages for the RJ reference.
• quantileMclust() uses bisection line search method for numerically computing quantiles.

Changes in version 5.4.5

• Fixed warnings in Fortran calls raised by CRAN.

Changes in version 5.4.4

• Added classPriorProbs() to estimate prior class probabilities.
• Added BrierScore() to compute the Brier score for assessing the accuracy of probabilistic predictions.
• Added randomOrthogonalMatrix() to generate random orthogonal basis matrices.
• Partial rewriting of summary.MclustDA() internals to provide both the classification error and the Brier score for training and/or test data.
• Partial rewriting of plot.MclustDA() internals.
• Added dmvnorm() for computing the density of a general multivariate Gaussian distribution via efficient Fortran code.
• Added Wisconsin diagnostic breast cancer (WDBC) data.
• Added EuroUnemployment data.
• Fixed mismatches in Fortran calls.
• Bugs fix.

Changes in version 5.4.3

• Added website site and update DESCRIPTION with URL.
• Fixed a bug when checking for univariate data with a single observation in several instances. Using NCOL() works both for n-values vector or nx1 matrix.
• Fixed a bug when hcPairs are provided in the initialization argument of mclustBIC() (and relatives) and the number of observations exceed the threshold for subsetting.
• Fixed bugs on axes for some manual pairs plots.
• Renamed type = "level" to type = "hdr", and level.prob to prob, in surfacePlot() for getting HDRs graphs
• Fixed a bug in type = "hdr" plot on surfacePlot().
• Fixed a bug in as.Mclust().
• Small changes to summary.MclustDA() when modelType = "EDDA" and in general for a more compact output.

Changes in version 5.4.2

• Added mclustBICupdate() to merge the best values from two BIC results as returned by mclustBIC().
• Added mclustLoglik() to compute the maximal log-likelihood values from BIC results as returned by mclustBIC().
• Added option type = "level" to plot.densityMclust() and surfacePlot() to draw highest density regions.
• Added meXXI() and meXXX() to exported functions.
• Updated vignette.

Changes in version 5.4.1

• Added parametric bootstrap option (type = "pb") in MclustBootstrap().
• Added the options to get averages of resampling distributions in summary.MclustBootstrap() and to plot resampling-based confidence intervals in plot.MclustBootstrap().
• Added function catwrap() for wrapping printed lines at getOption("width") when using cat().
• mclust.options() now modify the variable .mclust in the namespace of the package, so it should work even inside an mclust-function call.
• Fixed a bug in covw() when normalize = TRUE.
• Fixed a bug in estepVEV() and estepVEE() when parameters contains Vinv.
• Fixed a bug in plotDensityMclustd() when drawing marginal axes.
• Fixed a bug in summary.MclustDA() when computing classification error in the extreme case of a minor class of assignment.
• Fixed a bug in the initialisation of mclustBIC() when a noise component is present for 1-dimensional data.
• Fixed bugs in some examples documenting clustCombi() and related functions.

Changes in version 5.4

• Model-based hierarchical clustering used to start the EM-algorithm is now based on the scaled SVD transformation proposed by Scrucca and Raftery (2016). This change is not backward compatible. However, previous results can be easily obtained by issuing the command: mclust.options(hcUse = "VARS") For more details see help("mclust.options").
• Added subset parameter in mclust.options() to control the maximal sample size to be used in the initial model-based hierarchical phase.
• predict.densityMclust() can optionally returns the density on a logarithm scale.
• Removed normalization of mixing proportions for new models in single mstep.
• Internal rewrite of code used by packageStartupMessage().
• Fixed a small bug in MclustBootstrap() in the univariate data case.
• Fixed bugs when both the noise and subset are provided for initialization.
• Vignette updated to include references, startup message, css style, etc.
• Various bug fixes in plotting methods when noise is present.
• Updated references in citation() and man pages.

Changes in version 5.3

• Added gmmhd() function and relative methods.
• Added MclustDRsubsel() function and relative methods.
• Added option to use subset in the hierarchical initialization step when a noise component is present.
• plot.clustCombi() presents a menu in interactive sessions, no more need of data for classification plots but extract the data from the clustCombi object.
• Added combiTree() plot for clustCombi objects.
• clPairs() now produces a single scatterplot in the bivariate case.
• Fixed a bug in imputeData() when seed is provided. Now if a seed is provided the data matrix is reproducible.
• in imputeData() and imputePairs() some name of arguments have been modified to be coherent with the rest of the package.
• Added functions matchCluster() and majorityVote().
• Rewrite of print and summary methods for clustCombi class objects.
• Added clustCombiOptim().
• Fixed a bug in randomPairs() when nrow of input data is odd.
• Fixed a bug in plotDensityMclust2(), plotDensityMclustd() and surfacePlot() when a noise component is present.

Changes in version 5.2.3

• Added native routine registration for Fortran code.
• Fixed lowercase argument PACKAGE in .Fortran() calls.

Changes in version 5.2.2

• Fixed a bug in rare case when performing an extra M step at the end of EM algorithm.

Changes in version 5.2.1

• Replaced structure(NULL, *) with structure(list(), *)

Changes in version 5.2

• Added argument x to Mclust() to use BIC values from previous computations to avoid recomputing for the same models. The same argument and functionality was already available in mclustBIC().
• Added argument x to mclustICL() to use ICL values from previous computations to avoid recomputing for the same models.
• Fixed a bug on plot.MclustBootstrap() for the "mean" and "var" in the univariate case.
• Fixed uncertainty plots.
• Added functions as.Mclust() and as.densityMclust() to convert object to specific mclust classes.
• Solved a numerical accuracy problem in qclass() when the scale of x is (very) large by making the tolerance eps scale dependent.
• Use transpose subroutine instead of non-Fortran 77 TRANSPOSE function in mclustaddson.f.
• Fixed predict.Mclust() and predict.MclustDR() by implementing a more efficient and accurate algorithm for computing the densities.

Changes in version 5.1

• Fixed slow convergence for VVE and EVE models.
• Fixed a bug in orientation for model VEE.
• Added an extra M-step and parameters update in Mclust() call via summaryMclustBIC().

Changes in version 5.0.2

• Added option to MclustBootstrap() for using weighted likelihood bootstrap.
• Added a plot method for MclustBootstrap objects.
• Added errorBars() function.
• Added clPairsLegend() function.
• Added covw() function.
• Fixed rescaling of mixing probabilities in new models.
• Bug fixes.

Changes in version 5.0.1

• Fixed bugs.
• Added print method for hc objects.

Changes in version 5.0.0

• Added the four missing models (EVV, VEE, EVE, VVE) to the mclust family. A noise component is allowed, but no prior is available.
• Added mclustBootstrapLRT() function (and corresponding print and plot methods) for selecting the number of mixture components based on the sequential bootstrap likelihood ratio test.
• Added MclustBootstrap() function (and corresponding print and summary methods) for performing bootstrap inference. This provides standard errors for parameters and confidence intervals.
• Added "A quick tour of mclust" vignette as html generated using rmarkdown and knitr. Older vignettes are included as other documentation for the package.
• Modified arguments to mvn2plot() to control colour, lty, lwd, and pch of ellipses and mean point.
• Added functions emX(), emXII(), emXXI(), emXXX(), cdensX(), cdensXII(), cdensXXI(), and cdensXXX(), to deal with single-component cases, so calling the em function works even if G = 1.
• Small changes to icl(), now it is a generic method, with specialized methods for Mclust and MclustDA objects.
• Fixed bug for transformations in the initialization step when some variables are constant (i.e. the variance is zero) or a one-dimensional data is provided.
• Changed the order of arguments in hc() (and all the functions calling it).
• Small modification to CITATION file upon request of CRAN maintainers.
• Various bug fixes.

Changes in version 4.4

• Added option for using transformation of variables in the hierarchical initialization step.
• Added quantileMclust() for computing the quantiles from a univariate Gaussian mixture distribution.
• Fixed bugs on summaryMclustBIC(), summaryMclustBICn(), Mclust() to return a matrix of 1s on a single column for z even in the case of G = 1. This is to avoid error on some plots.
• Moved pdf files (previously included as vignettes) to inst/doc with corresponding index.html.

Changes in version 4.3

• Fixed bug for logLik.MclustDA() in the univariate case.
• Added argument "what" to predict.densityMclust() function for choosing what to retrieve, the mixture density or component density.
• hc() function has an additional parameter to control if the original variables or a transformation of them should be used for hierarchical clustering.
• Added "hcUse" argument in mclust.options() to be passed as default to hc().
• Added the storing of original data (and class for classification models) in the object returned by the main functions.
• Added component hypvol to Mclust object which provide the hypervolume of the noise component when required, otherwise is set to NA.
• Added a warning when prior is used and BIC returns NAs.
• Fixed bugs in summary.Mclust(), print.summary.Mclust(), plot.Mclust() and icl() in the case of presence of a noise component.
• Fixed bug on some plots in plot.MclustDR() which requires plot.new() before calling plot.window().
• Fixed a bug in MclustDR() for the one-dimensional case.
• Corrections to Mclust man page.
• Various small bug fixes.

Changes in version 4.2

• Fixed bug in sim*() functions when no obs are assigned to a component.
• MclustDA() allows to fit a single class model.
• Fixex bug in summary.Mclust() when a subset is used for initialization.
• Fixed a bug in the function qclass() when ties are present in quantiles, so it always return the required number of classes.
• Various small bug fixes.

Changes in version 4.1

• Added icl() function for computing the integrated complete-data likelihood.
• Added mclustICL() function with associated print and plot methods.
• print.mclustBIC() shows also the top models based on BIC.
• Modified summary.Mclust() to return also the icl.
• Rewrite of adjustedRandIndex() function. This version is more efficient for large vectors.
• Updated help for adjustedRandIndex().
• Modifications to MclustDR() and its summary method.
• Changed behavior of plot.MclustDR(..., what = "contour").
• Improved plot of uncertainty for plot.MclustDR(..., what = "boundaries").
• Corrected a bug for malformed GvHD data.
• Corrected version of qclass() for selecting initial values in case of 1D data when successive quantiles coincide.
• Corrected version of plot BIC values when only a single G-component models are fitted.
• Various bug fixes.

Changes in version 4.0

• Added new summary and print methods for Mclust().
• Added new summary and print methods for densityMclust().
• Included MclustDA() function and methods.
• Included MclustDR() function and methods.
• Included me.weighted() function.
• Restored hierarchical clustering capability for the EEE model (hcEEE).
• Included vignettes for mclust version 4 from Technical Report No. 597 and for using weights in mclust.
• Adoption of GPL (>= 2) license.

Changes in version 3.5

• Added summary.Mclust().
• New functions for plotting and summarizing density estimation.
• Various bug fixes.
• Added clustCombi() and related functions (code and doc provided by Jean-Patrick Baudry).
• Bug fix: variable names lost when G = 1.

Changes in version 3.4.11

• Added NAMESPACE.

Changes in version 3.4.10

• Removed intrinsic gamma-

Changes in version 3.4.9

• Fixed hypvol() function to avoid overflow.
• Fixed hypvol() help file value description.
• Removed unused variables and tabs from source code.
• Switched to intrinsic gamma in source code.
• Fixed default warning in estepVEV and mstepVEV.

Changes in version 3.4.8

• Fixed output when G = 1 (it had NA for the missing z component).

Changes in version 3.4.7

• Removed hierarchical clustering capability for the EEE model (hcEEE).
• The R 2.12.0 build failed due to a 32-bit Windows compiler error, forcing removal of the underlying Fortran code for hcEEE from the package, which does not contain errors and compiles on other platforms.

Changes in version 3.4.6

• Added description of parameters output component to Mclust and summary.mclustBIC help files.

Changes in version 3.4.5

• Added densityMclust() function.

Changes in version 3.4.4

• Fixed bug in covariance matrix output for EEV and VEV models.

Changes in version 3.4.3

• Bug fixes.

Changes in version 3.4.2

• Moved CITATION to inst and used standard format
• BibTex entries are in inst/cite.
• Fixed bug in handling missing classes in mclustBIC().
• Clarified license wording.

Changes in version 3.4.1

• Corrected output description in mclustModel help file.
• Updated mclust manual reference to show revision.

Changes in version 3.4

• Updated defaultPrior help file.
• Added utility functions for imputing missing data with the mix package.
• Changed default max to number of mixture components in each class from 9 to 3.

Changes in version 3.3.2

• Fixed problems with \cr in mclustOptions help file

Changes in version 3.3.1

• Fixed plot.mclustBIC() and plot.Mclust() to handle modelNames.
• Changed "orientation" for VEV, VVV models to be consistent with R eigen() and the literature
• Fixed some problems including doc for the noise option.
• Updated the unmap() function to optionally include missing groups.

Changes in version 3.3

• Fixed bug in the "errors" option for randProj().
• Fixed boundary cases for the "noise" option.

Changes in version 3.2

• Added permission for CRAN distribution to LICENSE.
• Fixed problems with help files found by new parser.
• Changed PKG_LIBS order in src/Makevars.
• Fixed Mclust() to handle sampling in data expression in call.

Changes in version 3.1.10

• Added EXPR = to all switch functions that didn't already have it.

Changes in version 3.1.9

• Added pro component to parameters in dens() help file.
• Fixed some problems with the noise option.

Changes in version 3.1.1

• Default seed changed in sim*() functions.
• Added model name check to various functions.
• Otherwise backward compatible with version 3.0

Changes in version 3.1

• Most plotting functions changed to use color.
• Mclust() and mclustBIC() fixed to work with G=1
• Otherwise backward compatible with version 3.0.

Changes in version 3.0

• New functionality added, including conjugate priors for Bayesian regularization.
• Backward compatibility is not guaranteed since the implementation of some functions has changed to make them easier to use or maintain.