CHANGES IN VERSION 1.2.14 ------------------------------ NEW FEATURES o opls: no scaling option now available o plot.opls: when 'parAsColFcVn' is a character vector, it is converted to a factor (thus drawing ellipses by default) BUG FIXED o opls: computation of the matrix of correlations (when the number of features is <= 100) o plot.opls: handling of ellipse display when parEllipseL parameter set to NA (default) CHANGES IN VERSION 1.2.12 ------------------------------ NEW FEATURES o opls: multiclass PLS-DA implemented (PLS2 approach; comments added in the vignette) default number of permutations set to 20 (instead of 10) predictive components denoted in the tables by 'p' (instead of 'h' previously) OPLS(-DA): simplified "modelDF" data frame ('rotation' row has been suppressed) BUG FIXED o opls: PCA model with 'svd' and a single component CHANGES IN VERSION 1.2.11 ------------------------------ NEW FEATURES o opls: for OPLS(-DA), vipVn and orthoVipVn are now computed as the VIP4,p and VIP4,o described in Galindo-Prieto et al (2014) o plot.opls: changes in palette: black/grey colors for diagnostics and other colors for scores CHANGES IN VERSION 1.2.10 ------------------------------ BUG FIXED o plot.opls: minor bug fixed ('x-score' label and color display of test samples when 'subset' is not NULL) CHANGES IN VERSION 1.2.9 ------------------------------ NEW FEATURES o vignette: minor update CHANGES IN VERSION 1.2.8 ------------------------------ NEW FEATURES o vignette: "some words of warning" section added CHANGES IN VERSION 1.2.7 ------------------------------ NEW FEATURES o vignette: default figures from plot.opls included CHANGES IN VERSION 1.2.6 ------------------------------ NEW FEATURES o strF: object size now displayed in Mb (instead of bytes); minor corrections to handle all matrices and data frames (whatever the dimensions, mode, row and column names) o unit tests: new tests added to increase test coverage CHANGES IN VERSION 1.2.5 ------------------------------ NEW FEATURES o unit tests: warning message corrected in order to allow test coverage display on the Bioconductor page CHANGES IN VERSION 1.2.4 ------------------------------ NEW FEATURES o vignette: minor correction in the default number of permutations for single response (O)PLS(-DA) models CHANGES IN VERSION 1.2.3 ------------------------------ NEW FEATURES o vignette: detailed explanations of Q2Y computation CHANGES IN VERSION 1.2.2 ------------------------------ NEW FEATURES o plot.opls: 'plotVc' argument renamed 'typeVc' similarly to the default 'plot' function CHANGES IN VERSION 1.2.1 ------------------------------ NEW FEATURES o opls: now takes either a (numeric) data frame or matrix as 'x' input (instead of matrix only) o predict: now takes either a (numeric) data frame or matrix as 'newdata' input (instead of matrix only) CHANGES IN VERSION 1.1.11 ------------------------------ NEW FEATURES o NEWS format updated CHANGES IN VERSION 1.1.10 ------------------------------ NEW FEATURES o opls: PLS models can now be built for 'x' data with a single variable CHANGES IN VERSION 1.1.9 ------------------------------ NEW FEATURES o opls: default number of permutations set to 10 (instead of 100) as a compromise to enable both quick computation and a first hint at model significance o opls: maximum number of components in automated mode (predI = NA or orthoI = NA) set to 10 (instead of 15) o plot.opls: bug corrected in case of single component model without permutation testing CHANGES IN VERSION 1.1.8 ------------------------------ SIGNIFICANT USER-VISIBLE CHANGES o sacurine dataset name simplification: names of sample metadata are now 'age', 'bmi', and 'gender' instead of 'ageVn', 'bmiVn', and 'genderFc'; names of variableMetadata are now 'msiLevel', 'hmdb', 'chemicalClass' instead of 'msiLevelVn', 'hmdbVc', 'chemicalClassVc' CHANGES IN VERSION 1.1.7 ------------------------------ NEW FEATURES o Control to avoid overfitting was strenghtened by: i) setting the default number of permutations to 100 (instead of 0) and ii) changing the default plot ("summary") to include both the "permutation" and the "overview" graphics CHANGES IN VERSION 1.1.6 ------------------------------ NEW FEATURES o .sinkC argument added in the opls and plot.opls methods: Diversion of messages is required for Galaxy integration CHANGES IN VERSION 1.1.5 ------------------------------ NEW FEATURES o current implementation supports two-class only classification CHANGES IN VERSION 1.1.4 ------------------------------ NEW FEATURES o renamed slot of opls object: rotationMN -> weightStarMN CHANGES IN VERSION 1.1.3 ------------------------------ NEW FEATURES o Minor internal changes regarding default ellipse plotting options and checking the compatibility between the subset size and the cross-validation fold CHANGES IN VERSION 1.1.2 ------------------------------ BUG FIXES o Minor internal changes CHANGES IN VERSION 1.1.1 ------------------------------ SIGNIFICANT USER-VISIBLE CHANGES o The packaging was modified (but not the algorithms) to be consistent with other machine learning packages: 'opls' is now a class and the 'print', 'plot', 'predict', 'summary', 'fitted', 'coefficients' and 'residuals' methods are available (see the vignette) SIGNIFICANT USER-VISIBLE CHANGES o renamed method: roplsF -> opls o renamed arguments testVi -> now 'subset' which indicates the indices of the training (instead of the testing) observations o values: tMN -> scoreMN pMN -> loadingMN wMN -> weightMN bMN -> coefficients rMN -> rotationMN varVn -> pcaVarVn tOrthoMN -> orthoScoreMN pOrthoMN -> orthoLoadingMN wOrthoMN -> orthoWeightMN o new (S3) methods for objects of class 'opls' print summary plot predict fitted coefficients residuals CHANGES IN VERSION 1.0.2 ------------------------------ NEW FEATURES o roplsF outputs: '.char2numF' function (required for class prediction on new datasets); replaces 'c2nLs' CHANGES IN VERSION 1.0.1 ------------------------------ SIGNIFICANT USER-VISIBLE CHANGES o renamed roplsF arguments (to indicate integers more explicitly) 'predN' -> 'predI' 'orthoN' -> 'orthoI' 'crossvalN' -> 'crossvalI' 'permN' -> 'permI' 'parCompVn' -> 'parCompVi' 'parTopLoadN' -> 'parTopLoadI' CHANGES IN VERSION 0.99.7 ------------------------------ SIGNIFICANT USER-VISIBLE CHANGES o rename ('sacurine' data set) for compatibility with 'Workflow4Metabolomics.org': 'profileMN' -> dataMatrix 'sampleDF' -> sampleMetadata 'variableDF' -> variableMetadata CHANGES IN VERSION 0.99.6 ------------------------------ SIGNIFICANT USER-VISIBLE CHANGES o roplsF: handling of PLS models with single predictive component only CHANGES IN VERSION 0.99.4 ------------------------------ SIGNIFICANT USER-VISIBLE CHANGES o rename: 'sacuri' data set -> 'sacurine' CHANGES IN VERSION 0.99.2 ------------------------------ SIGNIFICANT USER-VISIBLE CHANGES o rename: ropF -> roplsF ropStrF -> strF o depreacated roplsF arguments: parObsColLabVcn: now available through the parLabVc and parAsColVcn arguments parObsLegVc: not useful o rename roplsF arguments: ncpN -> predN ncoN -> orthoN ploVc -> plotVc algC -> algoC savImg.pdfC -> fileFig.pdfC savLog.txtC -> fileInfo.txtC logL -> log10L perN -> permN scaC -> scaleC tesVi -> testVi vrbC -> verboseC parObsColVc -> parAsColVcn parPcsVn -> parCompVn parMahL -> parEllipsesL parObsLabVc -> parLabVc parMaiL -> parTitleL parNexN -> parTopLoadN NEW FEATURES o ropColF -> now split between .colorF and .legendF o ropMahF -> now .ellipseF