Generalized functional extended redundancy analysis.

Area of Science:

• Statistics
• Functional Data Analysis
• Generalized Linear Models

Background:

• Functional Extended Redundancy Analysis (FERA) integrates data reduction into functional linear models.
• Existing FERA methods are limited to linear models and do not accommodate various response data distributions.
• There is a need for methods that can handle diverse response types within a functional data framework.

Purpose of the Study:

• To extend Functional Extended Redundancy Analysis (FERA) to the framework of Generalized Linear Models (GLM).
• To develop a method capable of analyzing functional predictor data with various exponential-family response distributions.
• To provide a robust tool for describing predictor characteristics and summarizing predictor-response relationships in a broader statistical context.

Main Methods:

• The proposed method reduces each set of predictor functions to a single component.
• This component is then used for explaining exponential-family responses within the GLM framework.
• An iterative algorithm is developed to maximize a penalized log-likelihood criterion using a basis function expansion approach.

Main Results:

• Two simulation studies demonstrated the performance of the proposed method using synthetic data.
• The method successfully handles functional predictor data and various response distributions.
• Empirical application to two real-world examples showcased the method's practical utility.

Conclusions:

• The extension of FERA to GLMs provides a powerful new tool for functional data analysis.
• This approach accommodates a wider range of response variables, enhancing its applicability.
• The developed method offers a valuable approach for summarizing complex functional relationships in diverse data settings.