Subgroup detection in linear growth curve models with generalized linear mixed model (GLMM) trees

Area of Science:

• Statistics
• Biostatistics
• Longitudinal Data Analysis

Background:

• Growth curve models analyze response variable development over time.
• Subject heterogeneity is common and often requires explanation or prediction.
• Existing methods may lack accuracy or efficiency for complex growth patterns.

Purpose of the Study:

• To extend generalized linear mixed-effects model (GLMM) trees for longitudinal data analysis.
• To identify subgroups with different trajectories within linear growth curve models.
• To assess the performance of extended GLMM trees against other partitioning methods.

Main Methods:

• Extension of GLMM trees from clustered cross-sectional data to longitudinal data.
• Application to linear growth curve models.
• Performance assessment using simulated and real-world data, compared to LongCART and structural equation model (SEM) trees.

Main Results:

• Extended GLMM trees demonstrated higher accuracy than the original algorithm and LongCART.
• Performance was comparable to structural equation model (SEM) trees.
• GLMM trees handle discrete and continuous time series, are robust to random-effects specification, and offer faster computation.

Conclusions:

• Extended GLMM trees provide an accurate and efficient method for subgroup identification in growth curve analysis.
• This approach enhances the analysis of longitudinal data with heterogeneous trajectories.
• GLMM trees offer a flexible and computationally advantageous alternative to existing partitioning methods.