Changes in version 0.32 (2026-07-21)                  

New features

  - transitions() now reads inference results, not only fitted models.
    It accepts the output of bootstrap_lsa(), certainty_lsa(),
    stability_lsa(), permute_lsa(), compare_lsa(), bayes_compare_lsa()
    and lsa_lags(), and returns the same tidy one-row-per-transition
    data frame. significant = TRUE keeps the transitions the method
    itself flags, and sort = "strength" orders them by that method's
    signed effect. Previously the per-transition table of an inference
    result was reachable only through as.data.frame(), because the print
    methods show a header summary alone.
  - The transitions() generic is now transitions(fit, ...), so each
    class declares only the arguments backed by columns it actually has.
    Existing calls on a fit are unaffected.

Documentation

  - Vignettes restore the user's options() after changing digits, as
    CRAN requires.
  - Vignettes read every result with transitions() rather than coercing
    it with as.data.frame(), and select rows with the significant
    argument rather than subsetting.
  - Removed the citations, reference lists, DOIs, documentation URLs and
    linked author metadata that 0.31 added to the shipped vignettes.

                        Changes in version 0.31                         

CRAN candidate

  - Prepared the package for first CRAN submission.
  - Moved ggplot2 and cograph from Suggests to Imports so the plotting
    surface works out of the box; plotting examples now run during
    checks. The analytical core still depends only on base R.
  - Added a dedicated interoperability vignette covering wide data, long
    event logs, tna, Nestimate, cograph, and lsa_to_tna().
  - Added linked author metadata and Dynalytics framework links to all
    shipped vignettes.
  - Restored lsa_to_tna() for handing an lsa fit to tna tooling.
  - Added ingestion of Nestimate::build_network() netobjects through
    their prepared sequence data.
  - Stored the bundled engagement data as a data frame so
    tna::tna(engagement) works directly.
  - Removed dead package-site URLs from CRAN-visible metadata.
  - Added future CRAN installation instructions to the README.
  - Included NEWS.md in the source package.

                        Changes in version 0.3.0                        

Interoperability and documentation

  - Added integration tests for cograph, Nestimate, and the Dynalytics
    evidence surface.
  - Added native TNA-style aliases: weights = "tna" and weights =
    "relative" now map to transition probabilities.
  - Updated plotting documentation and vignettes to use weights = "tna"
    for probability-weighted transition networks.
  - Added and reorganised vignettes:
      - intro: conceptual overview and package map.
      - lagdynamics: concise quick start.
      - workflow: complete applied workflow.
      - interop: interoperability with sibling packages.
      - lag-transition-networks: transition-network interpretation.
      - confirmatory: evidence and uncertainty workflow.
      - plotting: plot gallery.
  - Removed public documentation references to unexported internals.
  - Made long-format input more flexible: action is the only mandatory
    long-format column, with optional actor, session, time, and order.
  - Added warnings for single-sequence bootstrap and permutation cases
    where the requested procedure has limited inferential meaning.

                        Changes in version 0.2.0                        

Confirmatory workflow and group comparison

  - Added the Dynalytics-style confirmatory evidence battery:
    certainty_lsa(), bootstrap_lsa(), reliability_lsa(),
    stability_lsa(), and permute_lsa().
  - Added group comparison with compare_lsa() and Bayesian group
    comparison with bayes_compare_lsa().
  - Added grouped lsa() fits through group = ..., with grouped methods
    for transitions(), nodes(), tests(), initial(), plotting,
    reliability, and comparison workflows.
  - Added tidy as.data.frame() methods for inference and comparison
    result objects.
  - Added the unified plotting surface: residual heatmaps, residual
    networks, TNA probability networks, chord diagrams, sunbursts,
    uncertainty forests, and group-comparison plots.
  - Added native transition and initial probabilities:
    transition_probabilities() and initial().
  - Added bundled long-format data for examples and tests.

                        Changes in version 0.1.0                        

Initial implementation

  - Created a from-scratch, clean-room implementation of lag sequential
    analysis for categorical event sequences.
  - Added the unified lsa() constructor and canonical sequence handling
    through lsa_data() and lsa_transitions().
  - Added five built-in engines: classical, two_cell, bidirectional,
    parallel_dominance, and nonparallel_dominance.
  - Added convenience wrappers: lsa_classical(), lsa_two_cell(),
    lsa_bidirectional(), lsa_parallel_dominance(), and
    lsa_nonparallel_dominance().
  - Added the pluggable engine registry: register_lsa_engine(),
    get_lsa_engine(), list_lsa_engines(), and unregister_lsa_engine().
  - Added tidy reading verbs: transitions(), nodes(), tests(),
    initial(), and summary().
  - Added multi-lag helpers with lsa_lags() and lag_profile().
  - Added structural-zero handling through loops = FALSE and arbitrary
    structural-zero matrices.
  - Added experimental transfer_entropy() for directed categorical
    information-flow analysis.
  - Kept runtime dependencies minimal: only base R packages are imported
    (grDevices, grid, stats, and utils).