Changelog

All notable changes to this project will be documented in this file. The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.7.4] - 2022-04-30

Added

• Add duration field to StepTwo class to record the execution time for the spline fitting procedure
• Add CNAME for apex domain https://powerly.dev served via GitHub Pages.

Fixed

• Fix missing import for mvnorm package (#11). Closes #5.
• Fix missing number of cores in Backend when more cores than available were requested (#12). Closes #2.
• Fix legend overlapping spline confidence bands for StepThree plot (#13). Closes #3.
• Fix recording and reporting of Method and step classes execution time (#15). The duration is now recorded in seconds. Closes #9.

Removed

• Remove dev branch from all GitHub workflows (#14). Switched to the GitHub flow. Closes #4.

[1.7.3]

Changed

• Update badges order in README.md and removed open issues badge.

Fixed

• Fix typos in NEWS file.

[1.7.2]

Added

• Add GitHub badges with latest release version and number of open issues.

Changed

• Update GGM estimation test to check if the estimation fails when variables with zero standard deviation are present in the generated data.
• Update GGM estimation to fail when the generated data contains at least one variable that has a standard deviation equal to zero (i.e., as a result of generating data with a sample size value that is too low).

Fixed

• Add tolerance (i.e., 0.0000001) for test checking whether the spline coefficients are estimated correctly.
• Fix test for the updating of the bounds of a Range instance to run only when the the 2.5th and 97.5th selected sample sizes are different quantities.
• Fix bug in GGM data generating test where the number of nodes to generate data for were incorrectly sampled.

[1.7.1]

Changed

• Update moved URL https://codecov.io/gh/mihaiconstantin/powerly to https://app.codecov.io/gh/mihaiconstantin/powerly based on comments of CRAN maintainer Uwe Ligges.

[1.7.0]

Added

• Add summary S3 method support for Validation class objects.

Changed

• Add new R version to the CI pipeline for the R CMD check job.
• Update R CMD check workflow to use check action from r-lib/actions.
• Add explicit .data and .env pronoun calls (i.e., from the rlang package) to variables used inside data-masking function (e.g., ggplot2::aes()) to avoid CRAN notes regarding missing bindings for global variables. See this question for more information: https://stackoverflow.com/q/9439256/5252007.
• Update plot functions to respect the signature of the S3 plot generic.
• Improve documentation and positioning of figures in the help pages.
• Create roxygen2 templates for documenting the public API and S3 methods.
• Extract plot() methods from R6 classes into standalone S3 methods.
• Update package start-up logo with message to welcome and encourage users to submit ideas for improving the package.
• Extend CI workflow to include branch dev.

Fixed

• Fix backend test where the cluster would contain an unexpected .Random.seed. The .Random.seed is caused by loading the bootnet package which in turn loads the snow package. The snow package uses the RNG within the .onLoad() function to draw from the uniform distribution and set a port number. The result of this operation is a .Random.seed in the .GlobalEnv. For more information see: https://stackoverflow.com/q/69866215/5252007 and https://github.com/SachaEpskamp/bootnet/issues/82.
• Fix preprint URL in package documentation.
• Fix heading typo in news file.

[1.6.1]

Changed

• Update line exclusions for covr::codecov().

[1.6.0]

Changed

• Add new Validation class plot to README.md file.
• Update GgmModel tests to vary various model estimation and data generation parameters.
• Update testthat::expect_equal to test with a tolerance of 1e-6 for Basis and Solver.
• Replace own GgmModel estimation with qgraph::EBICglasso().
• Add ECDF plot to Validation class.
• Import external functions to NAMESPACE via Roxygen2.
• Add more cool badges to README.md (e.g., CRAN version and check status).
• Fix missing missing period in package description in DESCRIPTION file.

Fixed

• Fix issue where GGM estimation test would fail due to a precision level set too high.

[1.5.2]

Fixed

• Update Description field in DESCRIPTION file to follow CRAN guidelines.

[1.5.1]

Fixed

• Update preprint link in powerly() documentation to use the \doi{} syntax as indicated by CRAN member Uwe Ligges.
• Update DESCRIPTION to start with capital letter as indicated by CRAN member Uwe Ligges.
• Fix typo in DESCRIPTION.

[1.5.0]

Added

• Add preliminary summary() support for Method and Validation objects.
• Update citation and references to the preprint available at psyarxiv.com/j5v7u.
• Add ggplot plots and saving support for StepOne, StepTwo, StepThree and Validation objects.

Changed

• Migrate from R base plots to ggplot2 and patchwork.
• Add more detailed Description in DESCRIPTION file as per comment by CRAN member Julia Haider.

Fixed

• Fix seed in StepOne unit test.
• Fix superfluous test fail for GGM model estimation. Restricted the unit test to compare the estimated edge weights up to 7 digits precision.

[1.4.0]

Added

• Add code coverage workflow based on action provided by usethis::use_github_action("test-coverage") and badge to README.md.

[1.3.0]

Fixed

• Fix cluster creation bug in Backend when the machine had only 1 or 2 cores. The previous version would result in an error when the machine contained only 1 or 2 cores. Now, when the machine has only 1 core, the cluster creation will fail, with a message that not enough cores are available. When the machine has 2 cores, all 2 cores are used. In all other cases, the number of cores used is given by the number of cores on the machine minus one.

Changed

• Add names and comments to workflow file R-CMD-check.yaml.
• Update README.md and package startup logo to only show major version.
• Update existing tests and added new ones.
• Add type check for method argument of validate() to ensure that only instances of Method class (i.e., produced by powerly() are passed).
• Add Backend stopping to on.exit() in powerly() to ensure the cluster is stopped no matter the execution status of the function.
• Improve logic around setting and resetting the number of cores when the cluster is created and stopped (i.e., also for adopted clusters).
• Correct version number and add changes in NEWS.md for 1.2.0.

[1.2.0]

Added

• Add CI workflow via usethis::use_github_action_check_standard().

Fixed

• Restrict tests to using a maximum of two cores to respect CRAN restriction.

Changed

• Hide package logo at startup if the R session is not interactive.
• Add more informative changelogs to NEWS.md for previous releases.

[1.1.1]

Fixed

• Add .remove_missing() to clear any NA values that may be present in .measures in StepOne class after the Monte Carlo procedure.

[1.1.1]

Fixed

• Fix broken URLs in documentation causing build warnings.

[1.1.0]

Added

• Add validate() public API to validate method results.
• Add Validation class to perform validation on a Method object.

Changed

• Add example in README.md for validation() and fixed images.
• Decrease legend font size in StepThree class.
• Rename from quantile to percentile in StepThree class plots.

[1.0.0]

Added

• Add powerly() public API to run the method.
• Add generate_model() public API to generate true models.
• Add package documentation for public API.
• Add plotting for all three method steps.
• Add option to save last two iterations in Method class.
• Add Method class to run the method steps iteratively.
• Add parallel support for Step 1 and Step 3 of the method.
• Add Backend class to manage parallel clusters.
• Add support for multiple solvers for estimating spline coefficients.
• Rewrite prototype using R6 OOP style into first stable release.

[0.2.0]

Removed

• Remove text feedback from each method step function.

[0.1.1]

Fixed

• Fix bug cased by initial candidate sample size ranges being too narrow.

[0.1.0]

Added

• Add preliminary support for the Gaussian Graphical Model via ggm object.
• Add preliminary support for two statistics via statistic.power() and statistic.mean().
• Add validate.recommendation() prototype for validating a sample size recommendation.
• Add run.method() prototype for running all method steps iteratively.
• Add run.step.1() prototype for Step 1 of the method.
• Add run.step.2() prototype for Step 2 of the method.
• Add run.step.3() prototype for Step 3 of the method.

Planned changes and ideas to consider

• Add tests for Validation class.
• Option to set seeds for the cluster via parallel::setclusterSetRNGStream
• Mechanism for performing replicable simulations using the method.
• Mechanism for extending the package with new models, measures and statistics.
• Tutorial for methodologists on how to extend the package.
• Tutorial to demonstrate how the internal API of the package can be used.
• Add UML class diagram to give an overview of the structure of the package.
• Add package website using the pkgdown generator.
• Switch to generating README.md file from README.Rmd via knitr.
• Bisectional algorithm for choosing an optimal starting range (i.e., not too wide, not too narrow).
• Switch to accelerated bootstrap CI in StepThree for better precision.
• Reuse Monte Carlo results from previous iterations if they fall within the updated range.
• Shiny application for running the method.
• Allow rerunning the validation procedure with a custom sample size.
  • If multiple custom sample sizes are used during the validation, then switch from the histogram to a violin plot.