LoginSign Up
← Back to all sparks
T

TrendLSW

INFRA · APIS
Velocity0.0

Wavelet Methods for Analysing Locally Stationary Time Series

github.com/euanmcgonigle/trendlsw ↗

Wavelet trend estimation tightens the defaults it shipped with.

time-serieswaveletsdefaultsplottingcran
◆Current state
TrendLSW estimates trend and evolutionary wavelet spectrum for locally stationary time series through a single TLSW() entry point. Its history since the first CRAN appearance is short and centres on defaults and plotting around that function, plus one dataset addition. The latest entry carries both the 1.0.4 and 1.0.3 notes in one body.
◆Where it's heading
The package has moved from getting onto CRAN to correcting the choices it launched with: the spectrum filter defaults were swapped to their trend counterparts, the plot.CI switch was removed in favour of inferring it from what was actually computed, and an example was shrunk to fit check timings. This is consolidation around a stable API rather than expansion.
◆Prediction
Further releases most likely continue tuning TLSW() defaults and plot behaviour; the entries show no work toward new estimators.

◆Recent moves

  1. 8mo ago

    Trend filter defaults replace spectrum defaults; plot.CI dropped

    A cumulative body: the CRAN-driven example shrink sits on top of the 1.0.3 notes, which swap the S.filter.number and S.family defaults to their trend counterparts and drop plot.CI in favour of plotting whichever interval was actually computed. Defaults and plot behaviour both change for existing users.

    View source ↗
  2. 2y ago

    New z.acc and z.labels datasets shipped with the package

    Ships z.acc and z.labels as package data, giving the documentation a worked series to demonstrate the trend and spectrum estimators on.

    View source ↗
  3. 2y ago

    Description field and plot.TLSW documentation fixes

    Description field and plot.TLSW documentation touch-ups immediately after the first CRAN appearance, with no behaviour change.

    View source ↗