Changes in version 0.1.0                        

📉 Multiscale family

  - Added multiscale_entropy() — multiscale entropy (MSE; Costa,
    Goldberger & Peng 2002), the first function in the previously-empty
    multiscale family. Direct C++ port of the counting core in
    PhysioNet's reference mse.c (Costa), validated to reproduce the
    compiled reference binary's output exactly (to its own displayed
    precision) on synthetic test data, both at the sample-entropy level
    and for the full multiscale sweep.
  - Added sample_entropy() — sample entropy (SampEn; Richman & Moorman
    2000), filling a previously-planned gap in the entropy family and
    serving as the building block multiscale_entropy() applies at each
    coarse-grained scale.

🎲 Simulate family

  - Added pmodel() — multifractal binomial cascade generator (Meneveau &
    Sreenivasan 1987), with optional Fourier-domain fractional
    integration (Davis et al. 1997). Clean-room R reimplementation of
    pmodel.m (Victor Venema). Validated against two RNG-independent
    structural properties: exact constant output at p=0.5, and exact
    mass conservation at every cascade level for any p.

🧭 Ported from mrpheus

  - Added perm_entropy(), petrosian_fd(), hjorth_parameters(), and
    num_zerocross() — four nonlinear/complexity features centralised
    from Lucas França's own mrpheus package (its AASM staging feature
    pipeline, itself a validated R/C++ port of the YASA/antropy Python
    feature set). Re-validated directly against antropy 0.2.2 during
    this port: exact match for perm_entropy (normalized value),
    petrosian_fd, and num_zerocross. hjorth_parameters agrees closely
    but not exactly by design — it uses Bessel-corrected (ddof=1)
    variance, matching mrpheus's/R's convention, rather than antropy's
    population-variance (ddof=0) convention; the difference shrinks as
    signal length grows. Rtractor's existing higuchi_fd() was
    cross-checked against mrpheus/antropy's version and found to agree
    exactly for realistic epoch lengths.

🌿 Fractal family

  - Added dfa() — Detrended Fluctuation Analysis. Direct C++ port of
    PhysioNet's reference dfa.c (Mietus, Peng & Moody), validated to
    reproduce the compiled reference binary's output exactly on
    synthetic test data. Package license changed to GPL (>= 2) as a
    result (see inst/COPYRIGHTS).
  - Added higuchi_fd() — Higuchi Fractal Dimension. Clean-room C++
    reimplementation from Higuchi (1988), validated against a MATLAB
    reference implementation on synthetic test data (max abs. difference
    ~1e-11).
  - Added mfdma() — multifractal detrending moving average (Gu & Zhou
    2010). Clean-room C++ reimplementation from the published algorithm,
    segment-fluctuation core validated against a Python transliteration
    of the reference MATLAB implementation on synthetic test data (exact
    match to displayed precision).
  - Added chhabra_jensen() — multifractal spectrum via the
    Chhabra-Jensen box-counting method (1989). Clean-room C++
    reimplementation, moments core validated against a Python
    transliteration of the reference implementation (exact match to
    displayed precision).

🔁 RQA family

  - Added recurrence_microstate_entropy() — recurrence microstates
    maximum-entropy threshold search (Corso et al. 2018; Prado et
    al. 2018). Direct C++ port of the deterministic core of
    Max_Entropy() from circadia-bio/maxEntropy (Entropy.jl, MIT
    licensed), the script used in Ferre et al. (2024, PLOS ONE). Random
    pair sampling moved to R for set.seed()-reproducibility; core loop
    validated against a Python transliteration of the original (exact
    match on synthetic test data).

🚀 Initial scaffold

  - Repository initialised: package skeleton (DESCRIPTION, .Rprofile,
    R/, src/, tests/testthat/) with no exported functions yet.
  - Added Rtractor colour palette (rtractor_palette(),
    rtractor_palettes()), discrete/continuous ggplot2 scales
    (scale_colour_rtractor(), scale_fill_rtractor(),
    scale_colour_rtractor_c(), scale_fill_rtractor_c()), and
    theme_rtractor().
  - Planned function families stubbed as file headers only (no
    implementations yet, pending reference code inventory): entropy,
    fractal, lyapunov, multiscale, rqa, embed.

🎨 Visual identity

  - Hex sticker logo (man/figures/logo.svg), matching the Circadia Lab
    ecosystem's visual conventions: a dense woven strange-attractor
    motif (a nod to "Rtractor" = R + attractor) on a dotted-paper
    background, with the Rtractor colour palette carried through the
    border and wordmark. Wordmark baked as vector outlines rather than
    live text, so it renders correctly regardless of font availability
    in whatever tool processes the SVG (favicon generators, etc.).
  - pkgdown site live at https://rtractor.circadia-lab.uk, with the full
    favicon set generated via pkgdown::build_favicons().

📚 Documentation

  - vignette("getting-started") — tour of every implemented family.
  - vignette("multifractal-methods") — mfdma() vs chhabra_jensen(),
    validated against known pmodel() ground truth rather than just
    checking both run without error.
  - vignette("entropy-and-complexity") — perm_entropy()/
    sample_entropy()/multiscale_entropy(), including the classic Costa
    et al. white-noise-vs-correlated-signal multiscale entropy
    demonstration.