Functional Data Analysis with Causation in Observational Studies: Covariate Balancing Functional Propensity Score for Functional Treatments

TL;DR

This paper introduces a novel covariate balancing approach for estimating causal effects of functional treatments in observational studies, addressing the challenge of functional data without a density function.

Contribution

It proposes the first causal inference method for functional treatments using a multivariate substitute for propensity scores and covariate balancing techniques.

Findings

Method achieves good covariate balance in simulations.

Accurately estimates causal effects of functional treatments.

Applied to study body shape's effect on adipose tissue.

Abstract

Functional data analysis, which handles data arising from curves, surfaces, volumes, manifolds and beyond in a variety of scientific fields, is a rapidly developing area in modern statistics and data science in the recent decades. The effect of a functional variable on an outcome is an essential theme in functional data analysis, but a majority of related studies are restricted to correlational effects rather than causal effects. This paper makes the first attempt to study the causal effect of a functional variable as a treatment in observational studies. Despite the lack of a probability density function for the functional treatment, the propensity score is properly defined in terms of a multivariate substitute. Two covariate balancing methods are proposed to estimate the propensity score, which minimize the correlation between the treatment and covariates. The appealing performance of the proposed method in both covariate balance and causal effect estimation is demonstrated by a simulation study. The proposed method is applied to study the causal effect of body shape on human visceral adipose tissue.