Changelog • PLPBenchmarks

PLPBenchmarks 2.2.0

Github Actions workflows added for package checks and test coverage.
Updates to the whole test suit to speed up testing.
Prediction problems are now offered as datasets e.g. call data("tasks") to load a data frame with the problems.
Package DatabaseConnector now moved to Imports.
New vignettes added.
Github Actions workflow added for building the website. The package now has a dedicated website.
Bug fix in argument checks in createBenchmarkCohorts().
Fixed how arguments are passed into dplyr functions to limit “Notes” in package checks (R CMD check)
Updated README.md

ExecutionTimes are now exported as part of performance evaluation when calling getBenchmarkModelPerformance(). The times are derived from PatientLevelPrediction::runPlp() and refer to the TotalExecutionTime estimated in the runPlp function.
Andromeda R package updated to version 1.0.0. This is a major change for Andromeda as it switches the backend to duckdb. Also, packages FeatureExtraction updated to their latest compatible versions.
Updates also to the PatientLevelPrediction R package to its latest version.

Added tests.
Updated renv files so that now all profiles use PatientLevelPrediction@v6.4.0 from CRAN.
Solved a bug when incremental folder is defined outside saveDirectory.
Vignette updates

All function arguments are now being checked for validity.
New function to create a database and a shiny app to view results.
Default profile now loads PatientLevelPrediction@v6.4.0 from CRAN.
Function createBenchmarkCohorts() now uses entirely the R package CohortGenerator to create the cohorts.
Function loadBenchmarkDesigns() has been changed to loadModelDesigns().
Function viewBenchmarkSettings() can now accept list/lists of covariateSettings object/s.
Cohort definitions in json and sql formats are now linked with their corresponding cohortId.
Updates of problem specification and analysis designs.