Smoothing and Mean-Covariance Estimation of Functional Data with a Bayesian Hierarchical Model

Area of Science:

• Statistics
• Functional Data Analysis
• Bayesian Inference

Background:

• Functional data, where observations are functions, are common in many fields.
• Existing methods often smooth functions individually, risking the loss of shared patterns.
• Simultaneous smoothing of all functional observations is an underexplored area.

Purpose of the Study:

• To develop a nonparametric Bayesian approach for simultaneous smoothing of functional data.
• To enable borrowing strength across observations for improved accuracy.
• To retain systematic patterns present across multiple functions.

Main Methods:

• Proposed a nonparametric Bayesian framework assuming independent Gaussian processes for functional observations.
• Utilized a hierarchical model with Gaussian process prior for the mean function and Inverse-Wishart process prior for the covariance function.
• Implemented automatic mean-covariance estimation within the posterior inference.

Main Results:

• The proposed method achieves superior smoothing accuracy compared to existing techniques.
• It demonstrates comparable results in mean-covariance estimation.
• Successfully preserves systematic patterns often missed by individual-curve smoothing.

Conclusions:

• The nonparametric Bayesian approach offers effective simultaneous smoothing for functional data.
• The hierarchical framework is flexible for various data characteristics (e.g., grid types, covariance structures).
• This method advances functional data analysis by preserving essential cross-functional information.