LoginSign Up
← Back to all sparks
L

Luminescence

INFRA · APIS
Velocity2.5

Comprehensive Luminescence Dating Data Analysis

github.com/r-lum/luminescence ↗

Luminescence is revisiting the statistical assumptions baked into its dose-response fits.

r-packageluminescence-datinggeochronologydose-responsebreaking-changes
◆Current state
The package provides luminescence dating analysis for the geochronology community — signal import from several instrument formats, dose-response curve fitting, and a large library of analysis routines. The current release changes how fitting weights are handled by default, on the reasoning that individual relative uncertainties do not in fact scale with dose as the previous approach assumed. Several releases in this window are historical versions backfilled into the feed, so their timestamps sit within minutes of each other and do not reflect when the work shipped.
◆Where it's heading
Two kinds of change dominate. The first is deliberate breaking changes made for correctness or consistency: the weighting rework, and import functions now uniformly appending the detector to record types so all formats behave as one always did. The second is a steady stream of new analysis functions contributed by the wider research community, covering crosstalk correction, spatial autocorrelation on grain discs, and incomplete bleaching models. Deprecated functions are being removed on a clear schedule rather than left indefinitely.
◆Prediction
Expect further consolidation of defaults that were inherited rather than chosen, since the weighting change is framed as correcting earlier reasoning rather than adding an option.

◆Recent moves

  1. 25d ago

    Reworks dose-response fitting weights, changing prior defaults

    Changes what the dose-response fitting weights argument accepts, moving off a normalised inverse standard error default that assumed uncertainties scale with dose. Both approaches remain available, but results from prior versions will not reproduce under the new default.

    View source ↗
  2. 4mo ago

    Fixes an uninitialised array and background overcounting

    Bug fixes including an uninitialised C++ work array caught by CRAN's valgrind runs and a background integral that was overcounted through overlapping sub-intervals. Corrections rather than changes in approach.

    View source ↗
  3. 5mo ago

    Import functions now append detector type to record names

    A breaking consistency change: every import function now appends the detector to the record type, matching what one reader had always done. It regularises data coming from BIN, PSL, and Daybreak files at the cost of breaking code that matched on the old names.

    View source ↗
  4. 7mo ago

    Adds crosstalk correction, Moran's I, and dose-response fitting

    A backfilled historical release carrying substantial community contributions: crosstalk correction, Moran's I and scatterplots for single-grain discs, and a model for incomplete heterogeneous bleaching. The dose-response fitting functions reworked in the current release were introduced here.

    View source ↗
  5. 7mo ago

    Fixes baSAR reporting, fading input checks, and plot regressions

    A backfilled patch release fixing incomplete aliquot reporting, a missing input check, and a plotting regression. Routine corrections, listed out of version order because of how these releases were imported.

    View source ↗
  6. 7mo ago

    Adds automatic background subtraction and record removal helpers

    A backfilled release adding helpers that replace hand-written loops, notably background subtraction that identifies curve pairs automatically. The pattern of turning common analyst workflows into single functions runs throughout the package's history.

    View source ↗