Dynamic prediction in functional concurrent regression with an application to child growth

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Functional Data Analysis

Background:

• Dynamic prediction using historical data is crucial in many studies.
• Traditional mixed effects models lack flexibility for complex subject-specific trajectories.
• Concurrent predictors complicate existing dynamic prediction methods.

Purpose of the Study:

• To propose a novel dynamic functional concurrent regression model.
• To address limitations in flexibility and concurrent predictors for dynamic prediction.
• To enable accurate prediction of future trajectories from irregularly measured data.

Main Methods:

• Developed a dynamic functional concurrent regression model.
• Utilized a mixed effects representation for parameter and trajectory inference.
• Applied the model to predict children's length using historical measurements.

Main Results:

• The proposed model demonstrated superior accuracy in estimation and inference over existing methods.
• Simulation studies confirmed the model's enhanced performance.
• The model effectively handles irregularly measured functional responses and predictors.

Conclusions:

• The dynamic functional concurrent regression model provides a more flexible and accurate approach to dynamic prediction.
• This method is particularly beneficial when dealing with concurrent predictors and irregularly sampled data.
• Open-source software is available to implement these advanced statistical methods.