NEWS

---

csemGT 1.0.0 (2026-06-01)

Initial release

• csem_gt() implements the three relative-error estimators of Brennan (1998) -- full, large_a, uncorrelated -- together with the closed-form absolute error variance (Brennan 1998, eq. 20).
• Closed-form analytical sampling variances and item-resampling bootstrap variances for all four estimators.
• Quadratic smoothing of conditional error variances on observed score, with optional exclusion of floor and ceiling cases.
• D-study extrapolation via n_items_D.
• S3 class csem with full method set: print, summary, plot, as.data.frame, by_score, coef.
• iowa_like synthetic dataset calibrated to the ITED Vocabulary Test.
• Two vignettes: tutorial introduction and worked examples.