Robust and flexible estimation of stochastic mediation effects: A proposed method and example in a randomized trial setting

Kara Rudolph, Oleg Sofrygin, Wenjing Zheng, Mark J. Van Der Laan

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Causal mediation analysis can improve understanding of the mechanisms underlying epidemiologic associations. However, the utility of natural direct and indirect effect estimation has been limited by the assumption of no confounder of the mediator-outcome relationship that is affected by prior exposure (which we call an intermediate confounder)- A n assumption frequently violated in practice. We build on recent work that identified alternative estimands that do not require this assumption and propose a flexible and double robust targeted minimum loss-based estimator for stochastic direct and indirect effects. The proposed method intervenes stochastically on the mediator using a distribution which conditions on baseline covariates and marginalizes over the intermediate confounder. We demonstrate the estimator's finite sample and robustness properties in a simple simulation study. We apply the method to an example from the Moving to Opportunity experiment. In this application, randomization to receive a housing voucher is the treatment/instrument that influenced moving with the voucher out of public housing, which is the intermediate confounder. We estimate the stochastic direct effect of randomization to the voucher group on adolescent marijuana use not mediated by change in school district and the stochastic indirect effect mediated by change in school district. We find no evidence of mediation. Our estimator is easy to implement in standard statistical software, and we provide annotated R code to further lower implementation barriers.

Original languageEnglish (US)
Article number20170007
JournalEpidemiologic Methods
Volume7
Issue number1
DOIs
StatePublished - Jan 1 2018
Externally publishedYes

Fingerprint

Randomized Trial
Mediation
Random Allocation
Mediator
Public Housing
Randomisation
Estimator
Cannabis
Statistical Software
Direct Effect
Software
Covariates
Baseline
Simulation Study
Robustness
Experiments
Alternatives
Estimate
Demonstrate
Experiment

Keywords

  • data-dependent
  • direct effect
  • double robust
  • indirect effect
  • mediation
  • targeted maximum likelihood estimation
  • targeted minimum loss-based estimation

ASJC Scopus subject areas

  • Epidemiology
  • Applied Mathematics

Cite this

Robust and flexible estimation of stochastic mediation effects : A proposed method and example in a randomized trial setting. / Rudolph, Kara; Sofrygin, Oleg; Zheng, Wenjing; Van Der Laan, Mark J.

In: Epidemiologic Methods, Vol. 7, No. 1, 20170007, 01.01.2018.

Research output: Contribution to journalArticle

@article{20f2d1036a4643a2b714d75315f17e8f,
title = "Robust and flexible estimation of stochastic mediation effects: A proposed method and example in a randomized trial setting",
abstract = "Causal mediation analysis can improve understanding of the mechanisms underlying epidemiologic associations. However, the utility of natural direct and indirect effect estimation has been limited by the assumption of no confounder of the mediator-outcome relationship that is affected by prior exposure (which we call an intermediate confounder)- A n assumption frequently violated in practice. We build on recent work that identified alternative estimands that do not require this assumption and propose a flexible and double robust targeted minimum loss-based estimator for stochastic direct and indirect effects. The proposed method intervenes stochastically on the mediator using a distribution which conditions on baseline covariates and marginalizes over the intermediate confounder. We demonstrate the estimator's finite sample and robustness properties in a simple simulation study. We apply the method to an example from the Moving to Opportunity experiment. In this application, randomization to receive a housing voucher is the treatment/instrument that influenced moving with the voucher out of public housing, which is the intermediate confounder. We estimate the stochastic direct effect of randomization to the voucher group on adolescent marijuana use not mediated by change in school district and the stochastic indirect effect mediated by change in school district. We find no evidence of mediation. Our estimator is easy to implement in standard statistical software, and we provide annotated R code to further lower implementation barriers.",
keywords = "data-dependent, direct effect, double robust, indirect effect, mediation, targeted maximum likelihood estimation, targeted minimum loss-based estimation",
author = "Kara Rudolph and Oleg Sofrygin and Wenjing Zheng and {Van Der Laan}, {Mark J.}",
year = "2018",
month = "1",
day = "1",
doi = "10.1515/em-2017-0007",
language = "English (US)",
volume = "7",
journal = "Epidemiologic Methods",
issn = "2194-9263",
publisher = "Walter De Gruyter",
number = "1",

}

TY - JOUR

T1 - Robust and flexible estimation of stochastic mediation effects

T2 - A proposed method and example in a randomized trial setting

AU - Rudolph, Kara

AU - Sofrygin, Oleg

AU - Zheng, Wenjing

AU - Van Der Laan, Mark J.

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Causal mediation analysis can improve understanding of the mechanisms underlying epidemiologic associations. However, the utility of natural direct and indirect effect estimation has been limited by the assumption of no confounder of the mediator-outcome relationship that is affected by prior exposure (which we call an intermediate confounder)- A n assumption frequently violated in practice. We build on recent work that identified alternative estimands that do not require this assumption and propose a flexible and double robust targeted minimum loss-based estimator for stochastic direct and indirect effects. The proposed method intervenes stochastically on the mediator using a distribution which conditions on baseline covariates and marginalizes over the intermediate confounder. We demonstrate the estimator's finite sample and robustness properties in a simple simulation study. We apply the method to an example from the Moving to Opportunity experiment. In this application, randomization to receive a housing voucher is the treatment/instrument that influenced moving with the voucher out of public housing, which is the intermediate confounder. We estimate the stochastic direct effect of randomization to the voucher group on adolescent marijuana use not mediated by change in school district and the stochastic indirect effect mediated by change in school district. We find no evidence of mediation. Our estimator is easy to implement in standard statistical software, and we provide annotated R code to further lower implementation barriers.

AB - Causal mediation analysis can improve understanding of the mechanisms underlying epidemiologic associations. However, the utility of natural direct and indirect effect estimation has been limited by the assumption of no confounder of the mediator-outcome relationship that is affected by prior exposure (which we call an intermediate confounder)- A n assumption frequently violated in practice. We build on recent work that identified alternative estimands that do not require this assumption and propose a flexible and double robust targeted minimum loss-based estimator for stochastic direct and indirect effects. The proposed method intervenes stochastically on the mediator using a distribution which conditions on baseline covariates and marginalizes over the intermediate confounder. We demonstrate the estimator's finite sample and robustness properties in a simple simulation study. We apply the method to an example from the Moving to Opportunity experiment. In this application, randomization to receive a housing voucher is the treatment/instrument that influenced moving with the voucher out of public housing, which is the intermediate confounder. We estimate the stochastic direct effect of randomization to the voucher group on adolescent marijuana use not mediated by change in school district and the stochastic indirect effect mediated by change in school district. We find no evidence of mediation. Our estimator is easy to implement in standard statistical software, and we provide annotated R code to further lower implementation barriers.

KW - data-dependent

KW - direct effect

KW - double robust

KW - indirect effect

KW - mediation

KW - targeted maximum likelihood estimation

KW - targeted minimum loss-based estimation

UR - http://www.scopus.com/inward/record.url?scp=85053811530&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053811530&partnerID=8YFLogxK

U2 - 10.1515/em-2017-0007

DO - 10.1515/em-2017-0007

M3 - Article

AN - SCOPUS:85053811530

VL - 7

JO - Epidemiologic Methods

JF - Epidemiologic Methods

SN - 2194-9263

IS - 1

M1 - 20170007

ER -