BIMAM

BIMAM performs Bayesian imputation and joint analysis of missing variables across multiple studies to enable multilevel multiple imputation that accounts for study-level heterogeneity in epidemiological datasets.

Key Features:

• Simultaneous Imputation and Analysis: Performs imputation and downstream analysis within a single Bayesian model.
• Handling of Systematic and Sporadic Missing Data: Accommodates systematically missing variables across studies and sporadically missing data within studies for binary and continuous variables.
• Multilevel Multiple Imputation: Implements multilevel multiple imputation to model between-study heterogeneity.
• Random Effects Modeling: Includes many random effects to account for study-level heterogeneity.
• Bayesian Computation Interfaces: Uses R2MultiBUGS and MultiBUGS for Bayesian computation.

Scientific Applications:

• Epidemiological data integration: Enables pooling and joint analysis of datasets from multiple studies with differing measurement sets to retain variables and studies that would otherwise be excluded.
• Analysis of heterogeneous multi-study datasets: Supports inference in settings with study-level heterogeneity and complex missingness patterns.

Methodology:

Bayesian framework implementing multilevel multiple imputation with simultaneous imputation and analysis, incorporation of multiple random effects to capture study-level heterogeneity, and computation via R2MultiBUGS and MultiBUGS.