clinicalsignificance 3.0.0

Breaking changes

• cs_get_summary() now returns raw numeric values instead of formatted character strings (e.g., 0.5 instead of "50.0%"). This facilitates downstream analyses and plotting (e.g., in sensitivity analyses). Formatting is now handled exclusively by the respective print() and summary() methods.
• Refactored print() and summary() methods to be lighter and more maintainable. The output no longer relies on the cli package, which is now used solely for messages and warnings. This is an internal change to improve code readability; calculation results remain unaffected.

New features

• All main functions (cs_anchor(), cs_percentage(), cs_distribution(), cs_statistical(), cs_combined()) now calculate group-wise statistics if the group argument is provided.
• Added new datasets (garden_club, trackmania, columbo) to demonstrate various analysis methods.

Minor improvements and fixes

• Fixed an issue where the summary table for combined approaches was not displayed correctly.
• All vignettes have been updated to reflect the latest changes.

clinicalsignificance 2.1.0

• I included the JSS article doi:10.18637/jss.v111.i01 as a reference.

clinicalsignificance 2.0.0

Breaking changes

• The former main function clinical_significance() is deprecated. The package now offers several different approaches to clinical significance analyses, namely:
  • cs_anchor() (Anchor-based approaches)
  • cs_percentage()(Percentage-change approaches)
  • cs_distribution()(Distribution-based approaches)
  • cs_statistical()(Statistical approaches)
  • cs_combined()(Combined approaches)

The function clinical_significance() used the combined approach (distribution-based and statistical). So the identical analysis can be carried out with cs_combined().

• The function check_cutoff() is deprecated.

• When method = "HLM, participants with three or more data points will be used. Before, participants with at least three data points (i.e., 4 or more) were used.

New features

• You can now specify a significance level at which you want the analysis to be based on with the significance_level argument. The default is significance_level = 0.05. Note that for this is the Phi max level when you set method = "HA" as outlined in their article.

• The package now offers to combined approaches: The classic combination of statistical and distribution-based approaches but also the combination of statistical and anchor-based approaches

• We included the function cs_get_model() to retrieve the hierarchical linear model that is fit during the distribution-based approach if the HLM method is requested.

• The hechler_2014 dataset was included to ensure reproducibility of the JSS article.

Minor improvements and fixes

• Change the whole package to use R’s native pipe operator |>
• Remove unnecessary dependency on magrittr
• Include new vignettes replacing the old ones
• Display correct percentages in summary tables
• Update package vignettes

clinicalsignificance 1.2.1

Breaking changes

• Remove default theming from check_cutoff() as in other plot functions

Minor improvements and fixes

• Improve documentation of datasets
• Include examples for summary() and plot() method

clinicalsignificance 1.2.0

Breaking changes

• Remove default theming from plot() to give the user more control over the overall appearance

Minor improvements and fixes

• Further implementation of tidyeval principles since aes_() was deprecated
• Small improvements in vignettes
• Better documentation for some functions

clinicalsignificance 1.1.0

• Fix HLM method empirical bayes estimation (#2). This now relies on the lme4 package
• Consider the instrument direction for group level summaries in the HA method (#1) which was not done in the initial release

clinicalsignificance 1.0.1

• Added a NEWS.md file to track changes to the package.
• Added package URLs to DESCRIPTION, including one for bug reports
• New package logo