Changes in version 1.21 msqrob 1.21.2 - msqrobLmer() stores the estimated variance component ratios of each lmer fit in getModel(model)$theta (lme4's parameterisation and names). - Failed fits (fitError) get NA instead of a log2 fold change of 0 in hypothesisTest() and topFeatures(). - msqrobLmer() no longer crashes on a feature without any observations; such features become fitError. - msqrobLmer() also accepts a data.frame as data. - Ridge models: the mean-model parameters keep the ridge prefix again (e.g. ridgeconditionB), so contrasts from createPairwiseContrasts(ridge = TRUE) and existing ridge contrasts work again (1.21.1 gave a log2 fold change of 0). - getContrast() and varContrast() give NA instead of 0 for contrasts on names that are no parameters of the model (e.g. a typo). - msqrobLmer() fits mixed models with a rank-deficient design (e.g. a missing reference level) instead of returning fitError; non-estimable contrasts are NA, as for msqrobLm(). - msqrobAggregate() (and msqrobLmer() with rowdata): when the formula uses exactly one sample variable and one feature variable (e.g. ~ condition + (1 | Sequence)), the sample variable lost its type: a factor became integer codes and a numeric covariate was treated as a factor. - msqrobHurdle(ridge = TRUE): intensity models are stored on the correct rows (when the rows were not sorted by name, proteins got the intensity model of another protein). - trend = "count" with random effects or ridge: the counts are linked to the correct features; missing counts give an error before fitting. - trend = "mean" no longer crashes when a feature has no observations. - trend = TRUE gives an informative error (use trend = "mean"). - msqrobLmer(): a failure after the lmer fit gives a fitError, no longer an incomplete lmer model. - No error, but a warning, when no feature can be fitted; failed fits (fitError) keep NA posterior variances and degrees of freedom. - Empirical Bayes variance moderation now always estimates the prior with limma's fitFDistUnequalDF1() (squeezeVar(legacy = FALSE)), using the fitted features only. Previously, limma used its moment estimator fitFDist() as soon as one feature failed to fit, and fitFDistUnequalDF1() otherwise, so the method depended on fit failures and filtering; the moment estimator can also give an infinite prior df when residual df are small. Posterior variances, and hence p-values, change slightly. Requires limma >= 3.62.0. msqrob 1.21.1 - Modular implementation of msqrob.R - Add estimability functionality to catch and solve issues with changing reference class - Adding unit tests for all functions in msqrob.R - Trended squeezevar using mean (limma trend) or using feature counts used for aggregation (DEqMS). msqrob 1.21.0 - Version bump Changes in version 1.20 msqrob 1.20.0 - Version bump Changes in version 1.19 msqrob 1.19.1 - Add functions to calculate log-normalisation factors that can be used with sweep function - Add function msqrobCollect to collect results tables - Add function plotVolcano to make volcanoplots for hypothesisTest tables - Add function createPairwiseContrasts to generate contrasts for all pairwise comparisons - Illustrate these new functionalities in the vignettes - New Vignettes for DIA (DIA-NN and Spectronaut) msqrob2 1.19.2 - Update cptac vignette (fread with argument check.names = TRUE) msqrob2 1.19.3 - Update vignettes (arguments fread check.names = TRUE and integer64 = "double") to avoid issues with readQFeatures - Update README to point to msqrob2book Changes in version 1.17 msqrob 1.17.2 (nothing yet) msqrob 1.17.1 - Fixed NEWS file - Fixed vignette upon renaming of longFormat to longForm msqrob 1.17.0 - New Bioconductor devel release 3.22 Changes in version 1.16 msqrob 1.16.0 - New Bioconductor stable release 3.21 Changes in version 1.15 msqrob 1.15.1 - fix: fixed fnames in cptac vignette and moved away from msdata to rely on MsDataHub instead msqrob 1.15.0 - New Bioconductor devel release 3.21 Changes in version 1.14 msqrob 1.14.1 - Added more flexibility in the msqrobLM function for models with missing variables - Fixed full rank check for models without ridge in certain cases msqrob 1.14.0 - New Bioconductor stable release 3.20 Changes in version 1.13 msqrob 1.13.1 - Implemented msqrobAggregate() method for SummarizedExperiment objects. msqrob 1.13.0 - New Bioconductor devel release 3.20 Changes in version 1.12 msqrob 1.12.0 - New Bioconductor stable release 3.19 msqrob 1.11.2 - Fixed issue related to levels of a ridge variable - Fixed issue when fitting only one random effect msqrob 1.11.1 - Fixed issues related to reference class changes in the models - Fixed issue related to colData assay levels - Refactored internal code msqrob 1.6.2 - Fixed issues related to residual degrees of freedom of overparameterized models. - Fixed issue relating to coldata variables as factors. msqrob2 1.5.4 - Removed the full lmer model that was accidently left in the statmodel object. Changes in version 1.5.3 - Added the option to use mixed models without ridge regression - Added the option to use rowdata variables in the mixed models Changes in version 1.5.1 - Fix weighted variance covariance matrix and QR decomposition in the msqrobLmer function Changes in version 1.1.1 - Fix filtering steps in vignette: now using QFeatures::filterFeatures() Changes in version 0.99.6 - Update authors in Description file Changes in version 0.99.5 - Fix standard errors on the model parameter estimates by msqrobLmer when using doQR = TRUE Changes in version 0.99.4 - Minor update vignette. Replace eval=FALSE in one R chunk so that the code is evaluated. Changes in version 0.99.3 - Resolved notes on signalers - Resolved examples with lines wider than 100 characters - Avoiding sapply and 1:... Changes in version 0.99.2 - Added a NEWS.md file to track changes to the package. - Changed dependency from R (>= 4.0) to R (>= 4.1) - Updated citation file Changes in version 0.99.1 - Submission to Bioc