Tolerance bands for functional data.

Area of Science:

• Biostatistics
• Functional Data Analysis
• Statistical Inference

Background:

• Biomedical research often requires estimating population ranges for individual observations.
• Univariate measurements use tolerance intervals, but functional data requires a different approach.
• Existing methods struggle with the complexities of functional data, especially with varying data density.

Purpose of the Study:

• To develop and propose a methodology for constructing tolerance bands for functional measurements.
• To extend the concept of tolerance intervals to functional data.
• To create methods applicable to both sparse and dense functional datasets.

Main Methods:

• Developed a methodology for constructing pointwise and simultaneous tolerance bands.
• Utilized functional principal component analysis within a mixed model framework.
• Employed bootstrapping to approximate necessary tolerance factors and account for decomposition uncertainty.

Main Results:

• The proposed methodology demonstrates generally acceptable performance for functional data analysis.
• Performance may be liberal with highly sparse and unbalanced datasets.
• The method effectively incorporates both sparse and dense functional data.

Conclusions:

• Tolerance bands provide a robust extension of tolerance intervals for functional data.
• The methodology is suitable for various biomedical applications with functional measurements.
• Further refinement may be needed for extremely sparse or unbalanced data scenarios.