Combining parametric and nonparametric models to estimate treatment effects in observational studies

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Causal inference in observational studies relies on accurate adjustment for confounding variables.
• Traditional models struggle with a high number of discrete-valued confounders due to data sparsity.
• This limitation hinders the reliable estimation of treatment effects in complex datasets.

Purpose of the Study:

• To propose a new, scalable model for estimating treatment effects in observational studies with high-dimensional confounders.
• To develop a method that combines parametric and nonparametric outcome models within a conjugate framework.
• To avoid computationally intensive methods like Markov chain Monte Carlo (MCMC).

Main Methods:

• A novel model integrating parametric and nonparametric outcome models.
• Conceptual data splitting to maintain a conjugate framework.
• Approximations using the central limit theorem and random sampling for scalability.

Main Results:

• The proposed method demonstrates competitive performance against benchmark models.
• Efficient computation is maintained even with high-dimensional confounders.
• Successful illustration on a large epidemiological health survey dataset.

Conclusions:

• The new model offers an efficient and scalable approach to causal inference in observational studies.
• It effectively addresses the challenges posed by high-dimensional confounding.
• The method shows promise for real-world applications in epidemiology and beyond.