IPTW and DR Estimators: Advancing towards the Adoption of Methodologies for Estimating Causal Treatment Effects in Observational Studies

Abstract

The propensity score refers to the likelihood of a subject being selected for intervention, given their observed baseline covariates. By assigning different weights to academic disciplines based on the reciprocal probability of receiving treatment, there will be a synthetic sample where the assignment of treatment is not influenced by baseline factors that were measured. Utilizing inverse probability of treatment weighting (IPTW) through the propensity score enables people to assemble unbiased or objective estimates regarding average treatment effects. In order to estimate the causal relationship between an exposure and an outcome, the second notion of doubly robust estimation combines an outcome regression with a propensity score model. Only when the statistical model is appropriately stated can unbiased estimates be obtained using the propensity score approach and outcome regression separately. However, even if only one of the two models is correctly specified, an unbiased estimator for the effect can be obtained by employing the doubly robust estimator. This introduction to inverse probability of treatment weighting and doubly robust estimators includes conceptual overviews, an application to students’ performance in high school, as well as discussions based on the project.