← Back to all sparks
F

forecasting

ANALYTICS
Velocity0.0

R package forecasting by HIDDA — release notes from GitHub.

HIDDA.forecasting is a book chapter's reproducibility artifact, not a package under development.

forecastingepidemiologyreproducibilityvignettescranr package
Current state
HIDDA.forecasting accompanies a book chapter on forecasting infectious disease counts; its vignettes reproduce the results presented there using arima, prophet, glarma, hhh4contacts and scoringRules. The 1.0.0 release states this outright — it is the version used for the chapter, pinned to CRAN package versions as of July 2018. Every release since has been a vignette rebuild against newer R and dependency versions.
Where it's heading
The release pattern is maintenance on an eight-year cadence dictated entirely by the surrounding ecosystem: 1.1.1 rebuilt under R 4.0.4, 1.1.2 under R 4.3.2, 1.1.3 under R 4.6.1, each reporting whether the numbers moved. They mostly have not — the recurring note is minor numerical differences confined to the prophet forecasts in vignette('CHILI_prophet'). The only substantive change in the visible history is 1.1.0's methodological tidy-up of the scoring comparisons.
Prediction
Nothing in these entries points to new functionality; the next release is most likely another vignette rebuild whenever a dependency change or a CRAN check failure forces one.

Recent moves

  1. 1mo ago

    Vignettes rebuilt under R 4.6.1

    Vignettes rebuilt under R 4.6.1 with current dependency versions, producing only minor numerical differences in the prophet forecasts. Housekeeping to keep the book chapter's results reproducible on a current toolchain.

    View source ↗
  2. 2y ago

    Vignettes rebuilt under R 4.3.2

    The same rebuild pattern one cycle earlier, under R 4.3.2 on a new machine, again with minor differences only in the long-term prophet forecasts. Also accommodates checks run without suggested packages installed.

    View source ↗
  3. 5y ago

    Vignettes rebuilt under R 4.0.4

    A vignette rebuild under R 4.0.4 with no reported change in results. The entry carries no other content.

    View source ↗
  4. 7y ago

    Standard PIT and discretized log-normal scoring

    The one release in this history with methodological content: standard PIT replaces non-randomized PIT histograms for the continuous forecasters, and a new scores_lnorm_discrete() investigates discretized log-normal scoring. Both changes are reported as making almost no difference to the results, which is itself the finding.

    View source ↗
  5. 7y ago

    The version used for the book chapter, with pinned dependencies

    The reference release — the exact version used for the book chapter, with all dependency versions pinned to CRAN as of 25 July 2018 and named in the notes. It establishes the package's purpose as a frozen reproduction target rather than an evolving forecasting toolkit.

    View source ↗