UniCensorEM News

Version 0.1.0 (2026-07-18)

Initial Release

New Features

• Generalized EM algorithm for parameter estimation under censoring
• Support for 23 censoring schemes:
  • Complete data
  • Right censoring
  • Left censoring
  • Interval censoring
  • Random censoring
  • Block random censoring
  • Type-I censoring
  • Type-II censoring
  • Progressive Type-II censoring
  • Progressive first failure censoring
  • Joint Type-I censoring
  • Joint Type-II censoring
  • Balanced joint progressive Type-II censoring
  • Hybrid censoring
  • Hybrid Type-I censoring
  • Hybrid Type-II censoring
  • Type-I hybrid censoring
  • Type-II progressively hybrid censoring
  • Doubly Type-II censoring
  • Middle censoring
  • Right truncation
  • Left truncation

Core Functionality

• em_fit(): Main function for EM parameter estimation
• em.control(): Control parameters for EM algorithm
• dist_spec_em(): Define distribution for EM estimation
• Support for arbitrary user-supplied PDF, CDF, and survival functions
• Automatic construction of observed likelihood, complete likelihood, and Q-function
• E-step with conditional expectations for censored data
• M-step with multiple optimization methods (BFGS, L-BFGS-B, Nelder-Mead, CG, SANN, Brent, nlminb)

Statistical Inference

• Numerical Hessian computation for standard errors
• Bootstrap standard errors (parametric and nonparametric)
• Multiple confidence interval types (normal, percentile, BCa, basic)
• Louis method support (framework in place)

Model Selection

• AIC, AICc, BIC, HQIC, CAIC, KIC
• Takeuchi Information Criterion (TIC)
• model_selection() function for comprehensive criteria

Goodness-of-Fit

• KS statistic
• Anderson-Darling statistic
• Cramer-von Mises statistic
• Watson statistic
• RMSE, MAE, MSE
• R-squared and adjusted R-squared
• gof_stats() function

Diagnostics

• EM convergence diagnostics
• Parameter convergence tracking
• Log-likelihood convergence tracking
• Q-function convergence tracking
• Aitken acceleration support
• em_diagnostics() function

S3 Methods

• print.emfit(): Print fit summary
• summary.emfit(): Comprehensive summary
• plot.emfit(): Diagnostic plots (6 plot types)
• coef.emfit(): Extract parameter estimates
• vcov.emfit(): Variance-covariance matrix
• confint.emfit(): Confidence intervals
• logLik.emfit(): Log-likelihood
• AIC.emfit(): Akaike Information Criterion
• BIC.emfit(): Bayesian Information Criterion
• residuals.emfit(): Residuals
• fitted.emfit(): Fitted values
• predict.emfit(): Predictions

Residual Analysis

• Cox-Snell residuals
• Martingale residuals
• Deviance residuals
• Pearson residuals
• Score residuals
• Schoenfeld residuals
• Randomized quantile residuals
• Generalized residuals
• residual_analysis() function
• plot_residuals() function

Bootstrap

• Parametric bootstrap
• Nonparametric bootstrap
• Bias-corrected bootstrap
• Percentile bootstrap
• BCa intervals
• bootstrap_se() function

Visualization

• Parameter convergence plots
• Log-likelihood plots
• Q-function plots
• Histogram with fitted density
• QQ plots
• Residual plots
• Residuals vs fitted plots
• Residuals vs order plots

Parallel Computing

• Support for parallel package
• Support for future/furrr packages
• Parallel bootstrap
• get_ncores() function

Utilities

• Comprehensive error handling and validation
• Numerical integration utilities
• Automatic differentiation support (numDeriv)
• Safe numerical operations

Documentation

• Comprehensive roxygen2 documentation
• Benchmark examples vignette
• Examples for Exponential, Weibull, Gamma, Log-Normal, Normal distributions
• Unit tests using testthat

Performance

• Efficient numerical integration
• Vectorized operations
• Caching support
• Optional parallelization

Extensibility

• Modular censoring scheme architecture
• Easy to add new censoring schemes
• Distribution-agnostic core algorithm

Bug Fixes

• None (initial release)

Known Issues

• Some advanced censoring schemes use placeholder implementations
• Louis method requires further development
• Profile likelihood intervals not yet implemented

Deprecated

• None

Notes

• This package is designed as the estimation counterpart to UniCensor
• Together, they provide a complete framework for censored data analysis
• Compatible with R >= 4.0.0