ClusterBootstrap

ClusterBootstrap implements generalized linear models with cluster bootstrap (GLMCB) to provide robust estimation and inference for hierarchical or clustered data, including repeated measurements and longitudinal datasets with Gaussian and dichotomous outcomes.

Key Features:

• Assumption-free bootstrap: Employs cluster bootstrap resampling to reduce reliance on distributional and correlation-structure assumptions that underlie mixed models, generalized estimating equations (GEE), and ANOVA for repeated measurements.
• GLMCB implementation: Implements generalized linear models enhanced by cluster bootstrap (GLMCB) for parameter estimation and inference.
• Hierarchical data handling: Supports analysis of hierarchical data structures, including repeated measurements and longitudinal data.
• Flexible within-subject design: Accommodates non-factorial within-subject variables, unevenly spaced time variables, and time-varying covariates.
• Outcome types: Applicable to Gaussian and dichotomous dependent variables.
• Comparative evaluation: Performance has been compared to mixed models via Monte Carlo experiments, showing comparable or superior performance in some contexts.

Scientific Applications:

• Longitudinal and repeated-measures studies: Provides robust inference for studies with clustered or within-subject measurements over time.
• Analyses violating standard assumptions: Suited for scenarios where assumptions of mixed models, GEE, or repeated-measures ANOVA are difficult to verify or are unmet.
• Complex within-subject designs: Enables analysis of studies with non-factorial within-subject variables, uneven sampling times, or time-varying covariates.

Methodology:

Applies generalized linear models combined with cluster bootstrap resampling (GLMCB) for estimation and inference, with Monte Carlo experiments used for performance comparison to mixed models.