Multivariate functional response regression, with application to fluorescence spectroscopy in a cervical pre-cancer...

Area of Science:

• Multivariate statistics
• Biostatistics
• Functional data analysis

Background:

• Scientific studies often yield multivariate functional data from single subjects.
• Integrating information across these diverse measurements can uncover underlying mechanisms.
• Fluorescence spectroscopy in cervical pre-cancer research presents such a data challenge.

Purpose of the Study:

• To propose a novel multivariate functional response regression model.
• To analyze correlations between functional variables and multi-level data structures.
• To identify differential expressions in fluorescence spectroscopy for disease assessment.

Main Methods:

• A two-stage linear transformation: basis expansion and principal component analysis.
• A fully Bayesian approach for parameter sampling in transformed space.
• Functional tests for localizing effects and controlling error rates.

Main Results:

• The model effectively reduces intra- and inter-function correlations.
• Bayesian inference provides insights in the original data domain.
• Analysis of fluorescence data identified specific regions of differential expression.

Conclusions:

• The proposed model offers a robust framework for multivariate functional data.
• Identified differential expression regions can serve as potential biomarkers.
• This approach enhances prognosis and disease assessment in cervical pre-cancer studies.