View in EDS

What to do with a mediating and confounding variable that is partly a mediator and partly a confounder in causal inference.

Saved in:
Bibliographic Details
Title: What to do with a mediating and confounding variable that is partly a mediator and partly a confounder in causal inference.
Authors: Takahashi, Masayoshi1 (AUTHOR) mtakahashi615@g.chuo-u.ac.jp
Source: Communications in Statistics: Simulation & Computation. Oct2025, p1-24. 24p. 8 Illustrations.
Subject Terms: *EMPIRICAL research, *MONTE Carlo method, CAUSAL inference, CONFOUNDING variables, STATISTICAL bias, MULTIPLE imputation (Statistics)
Abstract: AbstractWhat variable to control for is of great importance if the goal of analysis is causal inference using observational data. The literature in causal inference is unanimous that we should control for a confounder while we should not control for a mediator that comes in the middle of the causal path. However, suppose that one variable is simultaneously a confounder and a mediator. In other words, some observations in this variable are observed before the treatment variable, while other observations in this same variable are observed after the treatment variable. Let us call this variable a mediating confounder. If we control for a mediating confounder, there will be bias from mediation; if we do not control for a mediating confounder, there will be an omitted variable bias. Therefore, this seems to be an unsolvable problem. This article proposes a method of multiple imputation with cold deck imputation, where we regard a mediating confounder as a special type of missing data. The results of Monte Carlo simulations present the evidence that the proposed method is indeed superior to the existing methods. Real data on economic and political variables are used to illustrate the usefulness of the proposed method. [ABSTRACT FROM AUTHOR]
Copyright of Communications in Statistics: Simulation & Computation is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Business Source Index
Description
Abstract:AbstractWhat variable to control for is of great importance if the goal of analysis is causal inference using observational data. The literature in causal inference is unanimous that we should control for a confounder while we should not control for a mediator that comes in the middle of the causal path. However, suppose that one variable is simultaneously a confounder and a mediator. In other words, some observations in this variable are observed before the treatment variable, while other observations in this same variable are observed after the treatment variable. Let us call this variable a mediating confounder. If we control for a mediating confounder, there will be bias from mediation; if we do not control for a mediating confounder, there will be an omitted variable bias. Therefore, this seems to be an unsolvable problem. This article proposes a method of multiple imputation with cold deck imputation, where we regard a mediating confounder as a special type of missing data. The results of Monte Carlo simulations present the evidence that the proposed method is indeed superior to the existing methods. Real data on economic and political variables are used to illustrate the usefulness of the proposed method. [ABSTRACT FROM AUTHOR]
ISSN:03610918
DOI:10.1080/03610918.2025.2571971