NEWS R Documentation

News for package smurf

Changes in version 1.1.3 (2022-05-06)

• Fix bug with ifelse and NULL when computing GLM penalty weights for a design matrix with full rank (#7).

Changes in version 1.1.2 (2021-04-17)

• Add rmarkdown to the Suggests field in the DESCRIPTION file to avoid problems with knitr version >= 1.32.

Changes in version 1.1.1 (2021-03-26)

• Remove 'LazyData' from DESCRIPTION file as no data directory is present.

Changes in version 1.1.0 (2021-03-12)

Major changes

• Only apply ridge penalty to relevant predictors when determining GLM penalty weights.

• Correctly determine if the design matrix has full rank.

• Optimal values of lambda, in examples and vignette, are re-determined after changes in computation of GLM penalty weights.

• Use C++14 (instead of C++11) which requires R version 3.4 or higher.

Minor changes:

• glmsmurf: Check that factors have at least two levels.

• Update Devriendt et al. (2021) reference.

Changes in version 1.0.7 (2020-11-24)

Changes in R-code:

• plot_lambda: Properly check whether the object x of class glmsmurf contains components related to the selection of the sparsity parameter lambda.

• Use named families for glmnet whenever possible as this is faster.

Changes in version 1.0.6 (2020-05-16)

Changes in R-code:

• Require at least version 4.0 of glmnet as this allows for more families to be used.

Changes in documentation:

• plot.glmsmurf, plot_reest.glmsmurf and plot_lambda.glmsmurf: Link to plot.default instead of plot (graphics) to avoid warnings on R-4.0.0.

Changes in version 1.0.5 (2020-03-19)

Changes in R-code:

• Use numerical tolerance that is also suitable on platforms without a long double.

• Use five-fold instead of two-fold cross-validation in tests.

Miscellaneous changes:

• Use rent dataset from catdata again as this package will no longer be archived.

• Use default date format for vignette date.

Changes in version 1.0.4 (2020-02-09)

Changes in R-code:

• Adapt test for input for prior weights to catch multiple warnings.

Documentation changes:

• Use only one core to select the optimal value of lambda in the example of plot_lambda.

Changes in version 1.0.3 (2020-02-07)

Changes in R-code:

• Add extra tests for proximal operators, number of cross-validation folds and prior weights.

Miscellaneous changes:

• Add rent dataset from catdata as this package is scheduled to be archived on CRAN on 14 February 2020.

• Correct README to display correct pipeline status on GitLab.

Changes in version 1.0.2 (2019-11-25)

Changes in R-code:

• Select first element of class(obj) when obj might be a matrix object as matrix objects will also inherit from class array in R 4.0.0 (and hence class(obj) will be of length > 1).

• Add an extra test for the output of a glmsmurf object where a Graph-Guided Fused Lasso is used.

Documentation changes:

• Update to roxygen2 version 7.0.0.

• Add empty first line to Rent_example2.R to avoid problems with roxygen2 version 7.0.0.

• Move example files from /inst to /inst/examples.

Changes in version 1.0.1 (2019-05-11)

Changes in R-code:

• glmsmurf: Improve handling of coefficient names.

Documentation changes:

• plot_lambda: Replace \dontrun by \donttest as requested by CRAN.

Changes in version 1.0.0 (2018-12-03)

Changes in R-code:

• p: change order of group and refcat arguments.

• glmsmurf: catch errors when computing the maximum value of lambda (#2).

• Fix bug in standardization when a continuous predictor is penalized with a Lasso or Group Lasso penalty (#4).

Documentation changes:

• Move examples for S3 methods into example for glmsmurf.

Miscellaneous changes:

• Remove maintainer field in DESCRIPTION as it is already set using Authors@R.

• Change GitLab URL in README.

• Update Devriendt et al. (2018) reference.

• Add reference to Devriendt et al. (2018) in DESCRIPTION.

• First release on CRAN.

Changes in version 0.4.1 (2018-09-25)

• First public release on GitLab.

Changes in version 0.4.0 (2018-09-24)

Changes in R-code:

• glmsmurf: Use "cv1se" to indicate selection of lambda using cross-validation with the one standard error rule. E.g. "cv.dev.1se" is renamed to "cv1se.dev".

Changes in documentation:

• glmsmurf: Add note that selected value of lambda for out-of-sample selection and cross-validation is not (always) deterministic.

Miscellaneous changes:

• Add continuous integration (CI) on GitLab.

• Add tests for plot, plot_lambda and summary functions.

• Add tests for elements of glmsmurf-class related to selection of lambda.

• General vignette update.