CausalMetaR: An R package for performing causally interpretable meta-analyses.

Area of Science:

• Epidemiology
• Biostatistics
• Data Science

Background:

• Traditional meta-analysis struggles with causal interpretability across diverse populations.
• Leveraging multi-source data for causal effect estimation in a target population is challenging.
• Existing methods may not yield causally valid estimates for specific target populations.

Purpose of the Study:

• To present the CausalMetaR R package for robust causal effect estimation from multi-source data.
• To enable causal inference in a defined target population (internal or external).
• To provide tools for estimating average and subgroup treatment effects.

Main Methods:

• Implementation of doubly robust and non-parametric efficient estimators.
• Support for data-adaptive methods, including machine learning techniques.
• Utilization of cross-fitting for estimating nuisance models (treatment and outcome).

Main Results:

• The CausalMetaR package offers robust and efficient causal effect estimation.
• It facilitates the estimation of average and subgroup treatment effects for target populations.
• The package supports advanced statistical techniques for improved accuracy and flexibility.

Conclusions:

• CausalMetaR enhances causal analyses by enabling the use of multi-source data.
• The package provides a practical tool for researchers aiming for causally interpretable meta-analytic results.
• It promotes more reliable causal inference in complex data synthesis scenarios.