Changes in version 1.0.0 (2026-07-28)                  

First public release; on CRAN since 2026-07-28.

Core

  - lst_data() declares variable roles (id, groups, sampling weight,
    scores, ability estimates paired with individual IRT standard
    errors, item responses with scoring keys, replicate weights,
    plausible values).
  - lst_table() builds Excel-pivot-style tables: free combination of row
    variables, column variables and cell statistics, with optional
    "Total" margins; results available as tidy long data (as_long()) or
    a formatted wide layout (as_wide()).
  - Statistics: st_mean(), st_sd(), st_prop_above(), st_level_prop(),
    st_quantile(), st_count()/st_wcount(), item p-values st_pvalue() and
    option distributions st_option_dist().
  - Every statistic reports a sampling standard error in every cell,
    including st_option_dist(): each option share is computed as the
    weighted mean of its 0/1 indicator, so it reuses the same linearized
    and replicate-weight variance paths as the other statistics.

Variance engines

  - Linearized sampling variance by default; Woodruff intervals for
    quantiles.
  - Measurement variance propagated from individual IRT standard errors;
    probabilistic classification (method = "prob") robust to
    misclassification near cut scores; empirical-Bayes correction =
    "latent" for WLE/ML estimates.
  - Replicate weights (rep_weights, methods fay/brr/jk1/jk2) for
    PISA/TIMSS-style designs, with prefix expansion such as "W_FSTR".
  - Plausible values (pv = list(math = "PV#MATH")) with Rubin
    combination, composing with replicate weights.

Interfaces and output

  - One-shot lst_run() driven by YAML/JSON text or an Excel
    configuration workbook (lst_config_template()); JSON Schema and an
    llms.txt API digest ship in inst/ for automated agents.
  - Importers for csv/tsv, Excel, SPSS/SAS/Stata (value labels
    preserved) and Winsteps/ConQuest person files.
  - Export to styled Excel workbooks, machine-readable JSON and
    standalone HTML reports, each with rule-based plain-language
    interpretation.