NEWS

---

coxmnar 0.1.0 (2026-07-31)

• Initial release implementing Cox proportional hazards regression with missing not at random (MNAR) failure indicators following Liu and Liu (2026) doi:10.1007/s11222-026-10857-1.
• Implements five estimators: Adjusted Imputation ("ai"), Adjusted Complete ("ac"), Missing at Random ("mar"), Complete-Case ("cc"), and Full-Data ("full").
• Joint maximum likelihood estimation of probability models for outcome and missingness mechanism.
• Nadaraya-Watson kernel estimation of missingness propensity with coordinate standardization and 5 kernel options.
• Asymptotic sandwich variance estimator and nonparametric bootstrap variance estimator.
• Convenience function coxmnar_compare() for multi-method comparison.
• Data simulation wrapper simulate_coxmnar_data() supporting arbitrary mechanisms.