This repository contains codes and pipelines associated with the article Doubly robust treatment effect estimation with missing attributes by Mayer et al. (2020).
A full pipeline for estimating treatment effects in the presence of missing attributes, i.e., incomplete confounders and covariates, is provided in pipeline_causal_inference_with_missing_attributes.Rmd.
This pipeline can be applied directly on a custom data set (the default is a simulated toy example), provided that it suits the format as follows:
X.na: confounders. A data.frame of size#observations x #covariates. With or without missing values.W: treatment assignment. A binary vector coded either with{0,1}or with{FALSE,TRUE}(representing{control,treatment}). Without missing values.Y: observed outcome. A numerical or binary vector (if binary, then coded with{0,1}). Without missing values.
The methodology has been applied on a medical question, the effect of the drug tranexamic acid on mortality among traumatic brain injury patients. The data used for this application is extracted from the Traumabase® registry. This registry is only available upon request. However we provide the code used to analyse the data and to estimate the ATE in this context in TranexamicAcid/ate_analysis_traumabase_example.Rmd.
Mayer, Imke, Erik Sverdrup, Tobias Gauss, Jean-Denis Moyer, Stefan Wager, and Julie Josse. 2020. "Doubly Robust Treatment Effect Estimation with Missing Attributes." Annals of Applied Statistics 14 (3): 1409–31.