Regression analysis of longitudinal data with random change point

Area of Science:

• Statistics
• Biostatistics
• Longitudinal Data Analysis

Background:

• Existing regression methods for longitudinal data with change points often limit analysis to continuous responses.
• Current approaches typically focus on change points affecting only the response or individual trajectory trends, not subject-specific variations.

Purpose of the Study:

• To develop a new joint modeling approach for longitudinal data accommodating subject-specific random change points.
• To address effect heterogeneity of covariates occurring before and after the change point.

Main Methods:

• Combines a generalized linear mixed-effects model for longitudinal response with a random change point.
• Integrates a log-linear regression model to handle the random change point.
• Employs a maximum likelihood estimation procedure for inference.

Main Results:

• The proposed method allows for subject-specific change points and varying covariate effects.
• Asymptotic properties of the estimators, distinct from standard results, are established.
• Simulation studies indicate the method's practical utility.

Conclusions:

• The novel joint modeling approach effectively handles longitudinal data with random, subject-specific change points and covariate heterogeneity.
• The method is validated through simulations and applied to COVID-19 data, showing practical applicability.