Simultaneous confidence bands for functional data using the Gaussian Kinematic formula.

Area of Science:

• Statistics
• Functional Data Analysis
• Computational Statistics

Background:

• Simultaneous confidence bands (SCBs) are crucial for reliable inference in functional data analysis.
• Existing methods often struggle with high-dimensional domains or non-Gaussian data.
• The Gaussian Kinematic formula of t-processes (tGKF) offers a potential framework for robust SCB construction.

Purpose of the Study:

• To develop a novel construction for SCBs applicable to functional parameters over arbitrary dimensional compact domains.
• To demonstrate the effectiveness of the tGKF method, particularly under non-Gaussian conditions and with limited data.
• To extend the methodology for discrete sampling with additive noise.

Main Methods:

• Utilizing the Gaussian Kinematic formula of t-processes (tGKF) for SCB construction.
• Leveraging central limit theorems (CLTs) for parameter estimation to ensure asymptotic precision.
• Applying scale space ideas from regression analysis for extensions to discrete data.
• Conducting simulation studies to compare the proposed method with existing techniques.

Main Results:

• The tGKF method provides asymptotically precise covering for SCBs, even for non-Gaussian processes, provided a CLT holds.
• The proposed tGKF construction outperforms state-of-the-art methods for SCBs of the population mean, especially with small sample sizes.
• The method demonstrates computational efficiency, even for domains exceeding one dimension.
• A Rademacher multiplier-t bootstrap approach showed comparable performance to the tGKF.

Conclusions:

• The tGKF offers a powerful and computationally efficient tool for constructing SCBs in functional data analysis.
• The method is robust to non-Gaussian data and performs well in small sample scenarios.
• The approach is applicable to diverse real-world problems, including diffusion tensor imaging and spatio-temporal data analysis.