AutoTraj: Automated Group-Based Trajectory Modeling Engine

🚀 Overview AutoTraj is a high-performance, open-source Python engine designed to replicate and accelerate the mathematical architecture of SAS PROC TRAJ for Group-Based Trajectory Modeling (GBTM). By utilizing Numba-driven Just-In-Time (JIT) C-compilation and fully vectorized analytical Jacobians, AutoTraj reduces exploratory model evaluation time from hours of manual macro loops to just a few seconds. 🛠 Key Features in V1.0-Beta Mathematical Parity: Replicates the exact Logit likelihood and observed information matrix inversion logic of PROC TRAJ. Smart Search Grid: Automates the combinatorial evaluation of polynomial grids for 1-8 groups. Mathematical Safeguards: Programmatically catches singular matrices and unidentifiable models by monitoring Hessian condition numbers—protecting against the "false significance" of overparameterized models. Missing Data Support: Includes Full Information Maximum Likelihood (FIML) for MAR data and an optional joint likelihood Dropout Model for MNAR attrition. Publication Suite: Instant generation of interactive Plotly visuals, adequacy metrics (AvePP, OCC, Relative Entropy), and exportable demographics tables. 🔬 Validation Note This release is intended for peer validation. While the mathematical core has been benchmarked against established software, we recommend users verify complex models against their historical SAS outputs. 📥 Getting Started git clone https://github.com/Thornwell16/gbtm_project.git cd gbtm_project pip install -r requirements.txt streamlit run app.py 🤝 Feedback & Contributing If you encounter discrepancies in parameter estimates or find bugs in the UI, please open an Issue or contact the lead developer: Donald E. Warden, PhD, MPH - dewarden@memphis.edu