Clustering Functional Data With Measurement Errors: A Simulation-Based Approach.

Area of Science:

• Statistics
• Data Science
• Biostatistics

Background:

• Functional data analysis is crucial in science but susceptible to measurement errors.
• These errors distort data structure, leading to inaccurate clustering outcomes.
• Existing methods often neglect measurement errors, compromising reliability.

Purpose of the Study:

• To propose a novel simulation-based approach for robust functional data clustering.
• To mitigate the impact of measurement errors on clustering accuracy.
• To provide more reliable clustering results in practical applications.

Main Methods:

• Estimating functional measurement error distributions using repeated measurements.
• Applying clustering to simulated data from the conditional distribution of true functional data.
• Adjusting for measurement errors to rectify the observed contaminated data.

Main Results:

• The proposed method demonstrates superior numerical performance compared to naive approaches.
• Simulations confirm improved clustering accuracy when measurement errors are addressed.
• Application to a childhood obesity study yielded more dependable clustering results.

Conclusions:

• The simulation-based method effectively addresses measurement errors in functional data clustering.
• This approach offers enhanced reliability for scientific data analysis.
• It holds significant potential for applications in public health and other fields.