A fast bootstrap algorithm for causal inference with large data

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Estimating causal effects from large datasets is crucial in research and industry.
• The traditional bootstrap method for standard errors and confidence intervals is computationally intensive for large data.
• Modern causal inference techniques further increase the computational burden of the bootstrap.

Purpose of the Study:

• Introduce a novel bootstrap algorithm, the causal bag of little bootstraps (CB দাব)
• Enhance computational efficiency for causal inference with large datasets.
• Ensure consistent estimates and reliable confidence interval coverage.

Main Methods:

• Developed the causal bag of little bootstraps (CB দাব) algorithm.
• Evaluated algorithm performance using simulation studies.
• Assessed bias, confidence interval coverage, and computational time.
• Applied the method to a large observational dataset (Women's Health Initiative).

Main Results:

• The CB দাব algorithm significantly improves computational efficiency compared to the traditional bootstrap.
• The proposed method yields consistent estimates and desirable confidence interval coverage.
• Simulation studies demonstrate the algorithm's effectiveness in bias and coverage.

Conclusions:

• The causal bag of little bootstraps is a computationally efficient and statistically sound method for causal inference with large datasets.
• This algorithm facilitates the evaluation of causal effects in complex, large-scale studies.
• The method was successfully applied to analyze the effect of hormone therapy on coronary heart disease.