ClusterBootstrap: An R package for the analysis of hierarchical data using generalized linear models with the cluster...

Area of Science:

• Statistics
• Biostatistics
• Data Science

Background:

• Clustered and hierarchical data analysis requires careful consideration of statistical technique assumptions.
• Traditional methods like mixed models, generalized estimating equations (GEE), and ANOVA for repeated measurements have significant limitations, including assumption violations and data handling issues.
• These limitations impact the validity and applicability of results in longitudinal and repeated measures studies.

Purpose of the Study:

• Introduce generalized linear models with the cluster bootstrap (GLMCB) as a robust method for analyzing hierarchical data.
• Demonstrate the practical application of the ClusterBootstrap R package with empirical examples for Gaussian and dichotomous outcomes.
• Compare the performance of GLMCB against mixed models through a Monte Carlo simulation study.

Main Methods:

• The study introduces the GLMCB technique, a bootstrap-based approach for hierarchical data analysis.
• An R package, ClusterBootstrap, is presented for implementing GLMCB.
• A Monte Carlo experiment is conducted to compare GLMCB with mixed models.

Main Results:

• The GLMCB technique is relatively assumption-free and demonstrates comparable or superior performance to GEE.
• Empirical examples showcase the ClusterBootstrap package's utility for various dependent variable types.
• The Monte Carlo experiment indicates GLMCB is a viable and promising alternative to mixed models for longitudinal data.

Conclusions:

• The GLMCB method, implemented through the ClusterBootstrap R package, provides a flexible and assumption-light approach to analyzing hierarchical and longitudinal data.
• This technique addresses limitations of traditional methods, offering a powerful alternative for researchers.
• The ClusterBootstrap package offers an accessible tool for applying GLMCB in statistical practice.