                        Changes in version 0.8.5                        

    o   Function fastLasso() is now exported for use by other developers.

                 Changes in version 0.8.4 (2026-01-20)                  

    o   C++ code now uses arma::span() to avoid ambiguity (as std::span() now
	exists as well).

    o   Cleaned up linker flags for C++ code.

                 Changes in version 0.8.3 (2025-12-09)                  

    o   C++ code no longer includes R.h when including RcppArmadillo.h.

    o   Replaced deprecated function aes_string() from package ggplot2 with
	aes() and .data pronoun.

                 Changes in version 0.8.2 (2025-10-02)                  

    o   Minor fixes in C++ code to avoid deprecated functions in Armadillo 15.0.

    o   Cleaned up unused variable in C++ code for sparseLTS().

                 Changes in version 0.8.1 (2024-06-30)                  

    o   Fix in C++ code for rlars() to avoid elementwise multiplication of row
	and column vector due to new version of RcppArmadillo being stricter
	regarding dimensions.

    o   Minor fix in help file for function coefPlot().

                 Changes in version 0.8.0 (2023-09-26)                  

    o   Removed deprecated functions.

    o   Added rstandard() method for standardized residuals rather than
	residuals(standardize = TRUE).

    o   Added DOI's to references in help files.

    o   Minor change in behavior of diagnosticPlot() for the regression
	diagnostic plot of sparse LTS results when the number of selected
	variables is large compared to the number of observations. Added
	clear documentation why this plot may fail in such a situation.

    o   Added metainformation on proteins, genes, and cancer cell lines to
	the NCI-60 cancer cell panel data.

    o   Bugfix of inconsistent argument names of the internal generic function
	getComponent() one of its methods.

                 Changes in version 0.7.4 (2023-01-18)                  

    o   Bugfix in C++ code: deprecated argument of insert_rows() and
	insert_cols() no longer used.

                 Changes in version 0.7.3 (2022-08-12)                  

    o   Bugfix in sparseLTS(): function calls for fitting the model on training
	sets (when selecting lambda via prediction error estimation) are now
	evaluated in the correct environment.

                 Changes in version 0.7.2 (2021-11-23)                  

    o   Added CITATION file.

    o   Added reference to JOSS paper in package help file.

    o   Added URL and BugReports fields to DESCRIPTION file.

                 Changes in version 0.7.1 (2021-10-13)                  

    o   Unit tests are now run automatically with R CMD check.

    o   Help file of sparseLTS() now correctly documents that arguments for
	selecting the optimal value of the penalty parameter are ignored if
	only one value of the penalty parameter is supplied.

                 Changes in version 0.7.0 (2021-09-17)                  

    o   Re-implemented plots and added functions setupCoefPlot(),
	setupCritPlot(), and setupDiagnosticPlot() to extract all
	the relevant information.

    o   Removed deprecated functions.

    o   Function partialOrder() for finding the order of the smallest or
	largest values of a vector is now exported.

    o   Added NCI-60 cancer cell panel data.

    o   Added coef(), fitted(), and residuals() methods for objects of class
	"perrySeqModel".

    o   Updated broken links in documentation.

                 Changes in version 0.6.1 (2019-12-14)                  

    o   Internal C++ functions for testing are now properly registered.

                 Changes in version 0.6.0 (2019-12-09)                  

    o   Method for function weights() is now used to retrieve robustness weights
	from objects of class "sparseLTS" (with argument type = "robustness").

    o   C++ functions are now properly registered.

    o   The alternative back end for sparse least trimmed squares from package
	sparseLTSEigen is no longer supported and can no longer be used.

                 Changes in version 0.5.1 (2016-01-08)                  

    o   Explicitly calling C++ function std::abs() rather than abs() to avoid
	clang warning.

    o   Correctly importing functions head() and tail() from package 'utils' and
	function devAskNewPage from package 'grDevices'.

                 Changes in version 0.5.0 (2014-04-30)                  

    o   Added functionality for (robust) groupwise least angle regression.

    o   Added TopGear car data.

    o   Diagnostic plots now allow to pass arguments to covMcd().

    o   Removed PCA step from data cleaning RLARS to consolidate code.

    o   Updated package dependencies.

                 Changes in version 0.4.0 (2013-12-10)                  

    o   sparseLTS() no longer uses subsampling algorithm in the special case of
	alpha = 1.

    o   sparseLTS() now has argument 'normalize' to specify whether the predictor
	variables should be normalized.

    o   sparseLTS() now computes objective function with coefficients for
	normalized data (if applicable).

    o   Most required packages are now imports rather than depends.

                 Changes in version 0.3.2 (2013-05-02)                  

    o   Bugfixes in sparseLTS() preventing errors for high-dimensional data.

                 Changes in version 0.3.1 (2013-04-16)                  

    o   rlars now uses perryFit() instead of perryTuning() for prediction error
	estimation.

    o   Bugfix in rlars() allowing the number of variables to be sequenced to be
	larger than half the number of observations.

    o   Bugfix in sparseLTS() in case of only one predictor variable.

    o   Added tests for C++ implementation of the lasso.

                 Changes in version 0.3.0 (2013-03-13)                  

    o   Redesign of the class structure.

    o   Redesign of how C++ back end is called.

    o   Functionality of sparseLTSGrid() now included in sparseLTS();
	sparseLTSGrid() is now a deprecated wrapper function.

    o   Restructured internal code for computing initial subsets for sparse LTS.

    o   rlars() now supports data cleaning RLARS, with an extra PCA step for
	high-dimensional data.

    o   New argument 's' in rlars() to select the steps along the sequence for
	which to compute submodels

    o   fortify() and diagnosticPlot() methods for class "seqModel".

    o   Bugfix in predict() method for "sparseLTS" if object was computed without
	intercept.

                 Changes in version 0.2.2 (2012-10-30)                  

    o   Bugfix in sparseLTS() for more stability of the results.

    o   Bugfix in winsorize(): weights are now correctly returned as vector for
	a matrix with only one column.

    o   Bugfix in diagnosticPlot(): previous setting of devAskNewPage() is now
	retained on exit.

                 Changes in version 0.2.1 (2012-09-09)                  

    o   Bugfix in rlars(): formula method now only adds function call and model
	terms if the default method returns an "rlars" object, not if only the
	sequence is returned.

    o   Bugfix in rlars(): argument cl is now preferred over argument ncores for
	parallel computing, as stated in the help file.

    o   Plots are no longer using the opts() function from package ggplot2, which
	is deprecated since ggplot2 version 0.9.2.

                 Changes in version 0.2.0 (2012-08-24)                  

    o   Graphics are now based on package ggplot2 instead of lattice.

    o   Prediction error estimation is now based on package perry instead of
	cvTools.

    o   Parallel computing for sparseLTS() now available via OpenMP.

    o   rlars() is now using C++ code for variable sequencing, including
	parallelization of certain tasks via OpenMP.  Further parallel
	computing is implemented on the R level via package parallel.

    o   sparseLTSGrid() and rlars() now allow model selection based on the
	prediction error.

    o   coef(), fitted(), residuals() and wt() methods now have argument
	'drop' to control whether to reduce the dimension if possible.

    o   Renamed components 'weight' and 'raw.weights' of sparse LTS models to
	'wt' and 'raw.wt', and renamed the accessor function accordingly to wt().

    o   Print methods for "sparseLTS" and "sparseLTSGrid" now only show non-zero
	coefficients by default; also added argument to print method for "rlars".

    o   sparseLTS() and sparseLTSGrid() now store the raw fitted values.

    o   Bugfixes in C++ code for sparseLTS() and fastLasso() to prevent memory
	related errors.
