Reproduce “Modeling Parkinson's Disease Progression Using Longitudinal Voice Biomarkers: A Comparative Study of Statistical and Neural Mixed-Effects Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ran Tong, Lanruo Wang, Tong Wang, Wei Yan
A comparative study evaluated statistical and neural mixed-effects models for predicting Parkinson's disease progression using longitudinal voice biomarkers, finding that Generalized Additive Mixed Models achieved superior predictive accuracy (MSE 6.56) over neural network models in small patient cohorts (N=42).
Key insights
- Generalized Additive Mixed Models (GAMMs) offer a superior balance of interpretability, stability, and predictive accuracy for small patient cohorts in longitudinal studies.
- Neural Mixed Effects models, despite their theoretical flexibility, are prone to severe overfitting when applied to small-sample longitudinal clinical datasets (N=42).
- The inherent structural assumptions of statistical mixed-effects models (e.g., linearity, smoothness) function as effective regularizers, contributing to their robustness in data-limited scenarios.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.