########## Versions 2.7.0 and 2.7.1 (April 2023) ################### Versions 2.7.0 and 2.7.1 implement the following changes: 1) Arguments named Idt (Interval-valued objects) by Sdt (Symbolic data objects) in generic functions that may be exported and extended to packages moddeling other types of symbolic data (ex: histogram or distritutional data). 2) The summary method for IdtMclust objects includes an additional argument indicating if class membership should shown observation by observation or class by clas (default) 3) Some previously unitiallized variables in the C code, are now properly initialized. ########## Version 2.6.1 (May 2021) ################### Version 2.6.1 adds aditional arguments to the plotInfCrt and pcoordplot methods in order to give the user some control over the placement and apparence of the respective legends. ########## Version 2.6.0 (April 2021) ################### Versions 2.6.0 implements the following changes: 1) A set of automated tests (using the facilities of the testthat package) is added to the MAINT.Data directory 2) The function IData is modifed in order to also accept: 2.1) A vector with the number of Micro Units by each interval-valued observation (when known) 2.2) Matrices without row names in its first argument 3) The DACrossVal function now displays also confusion matrices in addition to group specif and global estimates of classification errors. Furthermore, a new function, called ConfMat, displays and returns confusion matrices directly from factor vectors of original and prediction classes. 4) Anohter accessor method for the IData NbMicroUnits slot was added, allowing filling in (and not just reading, as in previus version) with new values 5) Methods cbind and rbind for IData objects are now defined as generic S4 methods that inherit from methods derived from the corresponding base S3 methods. This allows for the traditional base rbind and cbind methods to remain available after calling MAINT.Data, what was not guaranteed in previous versions. ########## Version 2.5.0 (September 2020) ################### Version 2.5.0 implements the following changes: 1) All remaining MAINT.Data specific S3 methods and classes were converted into S4 methods and classes. 2) The MAINT.Data definition of several generic methods (ex: sd, BIC, AIC, logLik) already defined in other packages has beed droped, avoiding unecessary masking. 3) A dim method was added to the IData class. 4) THe dependency on the sn package was changed, so that MAINT.Data always calls se functions when necessary, and it no longer copies its code in the R directory 5) Several new examples have been added to the manual, which was also corrected for some spelling mistakes. ########## Versions 2.4.0 and 2.4.1 (June 2020) ################### Versions 2.4.0 and 2.4.1 implement the following changes: 1) The linear algebra routines were changed so that the multicolinearity is now always checked beased on the l2-condition norm of correlation (instead of covariance) matrices, with an upper limit on numerical non-singular covariance matrix given by an argument (k2max, defaulting to 1e8) that may be changed by the user. 2) The IData class and AgrMcDt functions were modified in order to accept inputs from any class that inherits from data.frames. 3) The IData class was extended with the slot NbMicroUnits that gives the number of micro-data units aggregated (when applicable) in each statistical IData unit. 4) The C++ linear algebra computations are now based on the RcppArmadillo package instead of being based on the RcppEigen package, as it was the case in previous versions. ########## Versions 2.2.0 throught 2.3.0 (May-August 2019) ################### Versions 2.2.0 through 2.3.0 implement the following changes: 1) The function AgrMcDt now prints an warning when degenerate interval were found in the microdata. 2) New accessor methods were added to class IData, in order to acces data frames with the MidPoints, LogRanges and Ranges without having to look directly of the IData slots. 3) A new class named IdtOutl was creasted, saving the information about the outliers found by the function getIdtOutl into an S4 object. This classes has corresponding print, summary, plot and several accessor methods for better inspection and visualization of its content. 4) A new function named pcoordplot was added. This function plots paralel coordinate plots, helping the interpretation of a model based clustering analysis of interval-valued data. 5) The function DACrossVal now prints automatically its error rate estimates every time it is run. 6) New accessor methods were added to class IdtMclust, in order to acces the results of a model based clustering analysis of interval-valued data, without having to look directly of the IdtMclust slots. 7) A new function named MANOVAPermTest was added. This function implements a permutation test in order to find p-values of Interval-valued MANOVA analysis on small samples, where the asymptotic chi-square approximation is not reliable. ########## Versions 2.1.0 and 2.1.1 (April 2019) ################### Versions 2.1.0 and 2.1.1 implement the following changes: 1) Improved plot methods and new extractors (e.g. colnames, rownames, MidPoints, LogRanges, Lbounds, Ubounds, for objects of class IData). 2) The value returned from function DACrossVal when argument LOO is set to FALSE now returns N.A. (instead of 0.) for combinations of folders and classes where no observation was evaluated. Furthemore, when LOO is set to TRUE DACrossVal now returns by default a vector with the estimates of the error rate by class, and not a three-dimensional array with a the individual results for each combination of observation and class (which can still be returned with LOO=TRUE, but now only if the argument fullresults is also set to TRUE). ########## Version 2.0.0 (September 2018) ################### Version 2.0.0 implements the following major changes: 1) A new function (AgrMcDt) for the cretation of IData objectos from the agregation of microdata. 2) The IData class was extended with plot, rbind and cbind methods. In additional several small bugs were corrected and algorithmic aspects of the implementation were improved. ########## Version 1.2.3 (August 2018) ################### Version 1.2.3 fixes a small bug in the interface of getIdtOutl. In previous versions the getIdtOutl argument IdtE was wrongly assumed to be a list instead of an object of class IdtSngNDE (single distribution classical normal estimators) or IdtSngNDRE (single distribution robust normal estimators), as required. This problem prevented getIdtOutl from automatically recovering the list of observations used in the re-weighted MCD estimator, from the RewghtdSet slot (when not NULL) of the IdtE argument, as originally intended. The problem was fixed in version 1.2.3 ########## Version 1.2.2 (July 2018) ################### Version 1.2.2 fixes a bug in function fasttle that occurred when this procedure was called with argument getalpha="TwoStep” and in the first step no outlier was detected. In this situation versions 1.2.1 and older failed to distinguish between objects of class IdtSngNDE (single distribution classical normal estimators) and IdtSngNDRE (single distribution robust normal estimators). The problem was fixed in version 1.2.2 ########## Version 1.2.1 (January 2018) ################### Version 1.2.1 fixes a bug in function getIdtOutl ########## Version 1.2.0 (November 2017) ################### The following changes were made in version 1.2.0 of R package MAINT.Data: 1) A new IData method, Idtmclust, implements model-based clustering of Iinterval Data based on Gaussian mixtures. 2) A bug introduced in version 1.1.4 for the fasttle and fulltle methods was corrected 3) The show method for objects of class IdtMANOVA (and classes derived from IdtMANOVA) was improved ########## Version 1.1.4 (October 2017) ################### The following changes were made in version 1.1.4 of R package MAINT.Data: 1) The IdtSngNDRE and IdtMxNDRE classes were expanded in order to include slots with the Robust Mahalanobis distances, the trimmed subsests, and other important information used in the robust estimation. We note that in previous versions this information was provided in a list, but only when otpType argument was not at its (now decrepated) default of "OnlyEst". 2) The file Makevars was changed in order to make it compatible with the MAC osx-x86 Operating systems. 3) Some minor bugs from previous versions were fixed. ########## Version 1.1.3 (October 2017) ################### Version 1.1.3 adds a Makevars file (in order to compress the C code library), three new example data sets (LoansbyRskLvs_mimmaxDt, LoansbyRskLvs_qntlDt and LoansbyPurpose_mimmaxDt) and fixes some minor bugs from previous versions. ########## Version 1.1.2 (June 2017) ################### Version 1.1.2 updates and improves the documentation of MAINT.Data, and corrects a previously unnoticed bug on the C code of function fullte. ########## Versions 1.1.0 and 1.1.1 (June 2017) ################### The following changes were made in versions 1.1.0 and 1.1.1 of R package MAINT.Data: 1) The routines for robust estimation and outlier detection were improved and made more flexible. In particular, Version 1.1.o implements a novel enhanced method for finding bias correction factors for the re-weighted MCD estimator, and adds the function getIdtOutl, that tries to indentify outliers comparing robust Mahalanobis distances with quantiles either from asymptotic chi-square distribtuions, or from F and Beta approximations that, according to the literature may work better in finite samples. 2) The methods rownames and colnames were added to the class IData, and methods var, cor (returning respectivelly matrices of variance-covariance and correlation estimates) were added to all classes derived from the class IdtE. ########## Versions 1.0.0 and 1.0.1 (October 2016) ################### This is a major upgrade of the package with several new features and improvements of the existing code. The two main extensions of version 1.0.0 are: 1) The implementation of methods for estimation, testing and discriminant analysis based on the Skew-Normal distribution. For details see the documentation of the mle, MANOVA and snda methods and the IDtSNDE, IDtNandSNDE, IDtSngSNDE, IDtSngNandSNDE, IDtMxSNDE, IDtMxNandSNDE, IdtSNlocda, IdtLocSNMANOVA, IdtLocNSNMANOVA, IdtGenSNMANOVA and IdtGenNSNMANOVA classes. 2) The implementation of methods for the robust estimation of parameters of the guassian distribution, as well as the corresponding methods for robust linear and quadratic discriminant analysis. For details see the documentation of fasttle, fulltle, Roblda and Robqda methods and the IDtSngNDRE and IDtMxNRDE classes. Furthermore, among other new features, version 1.0.0 also includes specializations of the generic vcov method (returning estimates of the variance-covariance matrices for the parameter estimators used) for almost all (all except those based on robust methods) classes derived from IdtE, i.e. the classes containing results of model estimation. ########## Versions 0.5 and 0.5.1 (October 2015) ################### The following changes were made in Versions 0.5 and 0.5.1 of R package MAINT.Data: 1) Several functions from the default package ""stats"" were imported into the MAINT.Data namespace to ensure that they did not become undefined globals. 2) The strict dependecy on packages MASS and miscTools was restored by an Imports directive in the DESCRIPTION file (ensuring that these packages are always attached when installed) ########## Version 0.4 (May 2015) ################# The strict dependecy on packages MASS and miscTools was replaced by a Suggests directive ########## Version 0.3 (October 2014) ################# The following changes were made in Version 0.3 of R package MAINT.Data: 1) In the DESCRIPTION file the Depends directive for imported packages was repalced by an Imports directive 2) Lines with more 100 characters in the documentation files were split into two 3) The longer examples in the documentation files were replace by faster examples ########## Version 0.2 (October 2011) ################# The following changes were made in Version 0.2 of R package MAINT.Data: 1) The deprecated direct use of sd for data.frames was replaced by sapply(object,sd). 2) The original functions head, tail, coef, predict, stdEr, lda and qda were imported in order to properly recognizing them in the definition of methods for MAINT.Data classes. 3) the S3 method print for class summaryIData, and the S3 methods "[", "[<-" "==" and "!=" for class IData, were registered. 4) The function is.IData was replaced by is(object,"IData")