A hidden Markov model addressing measurement errors in the response and replicated covariates for continuous...

Area of Science:

• Biostatistics
• Computational Statistics
• Neuroscience

Background:

• Parkinson's disease (PD) progression tracking is crucial for patient management.
• Existing methods may not fully account for measurement errors and within-subject variability in longitudinal data.
• Voice analysis offers a promising, non-invasive biomarker for PD progression.

Purpose of the Study:

• To propose an advanced statistical model for analyzing longitudinal data, specifically motivated by Parkinson's disease progression.
• To develop a method that robustly handles measurement errors and within-subject variability in response variables.
• To provide a flexible framework applicable to various monotonic nondecreasing processes.

Main Methods:

• Development of an inhomogeneous hidden Markov model with a continuous state-space.
• Implementation within a Bayesian framework.
• Utilizing an efficient Markov chain Monte Carlo (MCMC) method for parameter estimation.

Main Results:

• The proposed model successfully addresses challenges like measurement error and covariate variability.
• Simulation studies demonstrate the model's performance and reliability.
• The model was applied to a real-world dataset for tracking Parkinson's disease progression.

Conclusions:

• The proposed inhomogeneous hidden Markov model provides a powerful tool for analyzing complex longitudinal data.
• This statistical approach enhances the understanding of disease progression, particularly for Parkinson's disease.
• The methodology is broadly applicable to other monotonic processes with similar data characteristics.