LoginSign Up
← Back to all sparks
C

cocoon

INFRA · APIS
Velocity0.0

Extract, Format, and Print Statistical Output

github.com/jeffreyrstevens/cocoon ↗

A statistics-formatting helper in maintenance mode, tracking R-devel one fix at a time

r-packagestatisticsreportingmaintenancereproducibility
◆Current state
cocoon formats statistical output for manuscripts, converting model and test objects into publication-ready strings. Its surface settled early: format_stats() is a generic that dispatches on object class, introduced in 0.1.0 to supersede the earlier format_corr() and format_ttest(), and extended in 0.2.0 to cover aov, lm, glm and the lme4 and lmerTest mixed-model families. The two releases since have been single-issue compatibility fixes against changes to wilcox.test() in R-devel.
◆Where it's heading
The package reached feature completeness for its stated job quickly and has been in maintenance since early 2025. Both 0.2.1 and 0.3.1 address the same upstream moving part - how wilcox.test() computes exact versus asymptotic distributions in development versions of R - which is the shape of a package whose own code is stable and whose risk lives entirely in what it wraps. Nothing in the recent entries points at new statistical object types.
◆Prediction
Further releases are likely to stay reactive, triggered by R-devel or dependency changes rather than by new formatting methods, unless a specific model class is requested.

◆Recent moves

  1. 1mo ago

    Test fix for wilcox.test changes in R-devel

    Adjusts a test case to exact = TRUE after R-devel changed how wilcox.test() computes the asymptotic approximation. The second fix in this same thread, and like the first it is test-suite maintenance rather than a change in what the package formats.

    View source ↗
  2. 11mo ago

    Suppress an R-devel warning in htest testing

    Adds exact = TRUE to a wilcox.test() call to silence a warning under r-devel. The first instance of the upstream-tracking pattern that 0.3.1 would repeat nine months later.

    View source ↗
  3. 1y ago

    format_stats gains regression and mixed-model methods

    Extends the format_stats() generic from correlations, t-tests and Bayes factors to aov, lm, glm and the lme4 and lmerTest mixed-model families. This is the release that made the generic worth dispatching on, covering most of what applied researchers actually report, and it marks the point after which the package went quiet.

    View source ↗
  4. 1y ago

    format_stats generic supersedes the per-test formatters

    Replaces the separate format_corr() and format_ttest() entry points with a single class-dispatching generic, and adds support for output from correlation::correlation(). The API shape the package has kept ever since, with 0.2.0 filling in methods rather than changing the design.

    View source ↗
  5. 1y ago

    Pre-release tag of the 0.1.0 contents

    An alpha tag published twelve seconds before 0.1.0 and carrying the same notes. It is the pre-release artifact of that release, not a separate set of changes.

    View source ↗