Changelog

Source: NEWS.md

JointODE 0.1.2

Major Features

• Comprehensive Visualization System: Implemented complete plotting framework with plot.JointODE() method
  • Model diagnostics: overview panels, biomarker/velocity trajectories, phase space diagrams
  • Survival analysis: Kaplan-Meier curves, hazard contributions
  • Residual diagnostics: standardized residuals, Q-Q plots
  • All plots built with ggplot2 for publication-ready graphics
• Initial State Optimization: Added iterative optimization for estimating initial biomarker and velocity values
  • Uses CppAD automatic differentiation for efficient gradient computation
  • Improves model convergence and parameter estimation accuracy
  • Available in both JointODE() and MarginalODE() functions
• Enhanced Prediction: Extended predict() method with improved covariate handling via LOCF (Last Observation Carried Forward)

Package Quality Improvements

• Fixed R CMD check warnings for CRAN compliance
  • Renamed src/utils.hpp to src/utils.h following R package conventions
  • Added .covrignore to exclude third-party libraries from test coverage
• Optimized documentation examples (72s → 1.5s, ~98% improvement)
• Fixed NEWS.md format for proper changelog display

JointODE 0.1.1

• Implemented subject-specific random effects on ODE acceleration parameters to account for population heterogeneity
• Added Laplace approximation for efficient posterior computation of random effects
• Integrated AGHQ (Adaptive Gauss-Hermite Quadrature) for accurate numerical integration
• Implemented CppAD automatic differentiation for efficient gradient computation

JointODE 0.1.0

• Joint modeling of longitudinal biomarkers and survival outcomes using ODEs
• Second-order differential equation formulation for biomarker dynamics
• EM algorithm for parameter estimation
• Parallel processing support for improved computational efficiency

---

Development Roadmap

Version 0.2.0 (Planned - Q2 2025)

Subgroup Heterogeneity

• JointODE_group(): Latent subgroup modeling
• Group-specific ODE parameters (κ_g, γ_g)
• Model selection via ICL and entropy criteria
• K-means initialization with stability analysis

Version 0.3.0 (Planned - Q4 2025)

Multiple Biomarkers

• JointODE_multi(): Multi-marker joint modeling
• select_biomarkers(): SIP-based variable selection
• Adaptive LASSO and group penalties
• Rcpp integration for performance

Version 0.4.0 (Planned - Q4 2026)

Machine Learning

• JointODE_nn(): Neural ODE backend
• Python bridge via reticulate
• Cloud deployment tools