NEWS
seasight 0.1.0
Submission-readiness update
- Correct seasonal-amplitude scaling: log models now report the
peak-to-trough seasonal-factor range in percentage points, level models use
the additive range relative to the median absolute level, and unavailable
components produce
NA rather than zero.
- Render unavailable QS diagnostics explicitly in the existence and
engine-choice cards instead of producing incomplete labels or errors.
- Surface SEATS model substitutions in the report summary and engine-choice
card. The candidates table now records a stable switch indicator alongside
the requested and operative model specifications.
- Add
comparison_mode = "incumbent_fixed" for ARIMA comparisons that retain
the incumbent transformation, fixed outliers, calendar terms and stored user
regressors. Candidate metadata and reports identify the comparison design and
regressor differences.
- Fix
include_easter = "auto" so X-13 tests and selects the Easter effect
instead of always imposing easter[15]. Candidate metadata and HTML reports
now show the Easter window actually retained by the fitted model.
- Make
sa_should_switch() compare the selected candidate directly with the
incumbent. Identical adjustments are kept, a change requires a material
comparable AICc improvement without diagnostic regression, and missing or
non-comparable evidence is reported explicitly.
- Freeze core user-facing API for the
0.1.x pre-submission stabilization cycle.
- Add contributor and governance documentation (
CONTRIBUTING.md, CODE_OF_CONDUCT.md).
- Add package citation metadata via
inst/CITATION.
- Add initial rOpenSci statistical-review standards mapping with a bronze-first target.
- Clarify positioning versus alternatives in README and getting-started vignette.
seasight 0.0.1
- Initial development release.