Entropy balancing for causal generalization with target sample summary information.

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Estimating average treatment effect (ATE) is crucial for causal inference.
• Covariate shift, where treatment effect modifiers differ between source and target populations, complicates ATE estimation.
• Existing methods often require individual-level target data, which is not always available.

Purpose of the Study:

• To develop a method for estimating target population ATE using source individual data and target summary data.
• To adjust for covariate shift without requiring individual target covariates.
• To improve the accuracy of ATE estimation in target populations with limited data.

Main Methods:

• A weighting approach is proposed, calibrating source sample weights using target population summary statistics.
• The method adjusts for covariate shift in treatment effect modifiers.
• Covariate balance between treated and control groups within the source sample is also pursued.

Main Results:

• The developed weighting approach effectively adjusts for covariate shift.
• The method demonstrates accurate estimation of target population ATE.
• Theoretical implications are supported by simulation studies and a real-data application.

Conclusions:

• The proposed weighting method offers a viable solution for ATE estimation when only summary data from the target population is available.
• This approach enhances causal inference in situations with covariate shift and limited target data.
• The method provides a robust tool for researchers in epidemiology and related fields.