Assessing the sensitivity of methods for estimating principal causal effects

Research output: Contribution to journalArticle

Abstract

The framework of principal stratification provides a way to think about treatment effects conditional on post-randomization variables, such as level of compliance. In particular, the complier average causal effect (CACE) - the effect of the treatment for those individuals who would comply with their treatment assignment under either treatment condition - is often of substantive interest. However, estimation of the CACE is not always straightforward, with a variety of estimation procedures and underlying assumptions, but little advice to help researchers select between methods. In this article, we discuss and examine two methods that rely on very different assumptions to estimate the CACE: a maximum likelihood ('joint') method that assumes the 'exclusion restriction,' (ER) and a propensity score-based method that relies on 'principal ignorability.' We detail the assumptions underlying each approach, and assess each methods' sensitivity to both its own assumptions and those of the other method using both simulated data and a motivating example. We find that the ER-based joint approach appears somewhat less sensitive to its assumptions, and that the performance of both methods is significantly improved when there are strong predictors of compliance. Interestingly, we also find that each method performs particularly well when the assumptions of the other approach are violated. These results highlight the importance of carefully selecting an estimation procedure whose assumptions are likely to be satisfied in practice and of having strong predictors of principal stratum membership.

Original languageEnglish (US)
Pages (from-to)657-674
Number of pages18
JournalStatistical Methods in Medical Research
Volume24
Issue number6
DOIs
StatePublished - Dec 1 2015

Fingerprint

Causal Effect
Compliance
Predictors
Ignorability
Principal Stratification
Restriction
Propensity Score
Joints
Treatment Effects
Randomisation
Maximum Likelihood
Random Allocation
Assignment
Likely
Research Personnel

Keywords

  • complier average causal effect
  • intermediate outcomes
  • non-compliance
  • principal stratification
  • propensity scores

ASJC Scopus subject areas

  • Epidemiology
  • Health Information Management
  • Statistics and Probability

Cite this

Assessing the sensitivity of methods for estimating principal causal effects. / Stuart, Elizabeth; Jo, Booil.

In: Statistical Methods in Medical Research, Vol. 24, No. 6, 01.12.2015, p. 657-674.

Research output: Contribution to journalArticle

@article{7ebef40fe21947e593af48b3802fe04d,
title = "Assessing the sensitivity of methods for estimating principal causal effects",
abstract = "The framework of principal stratification provides a way to think about treatment effects conditional on post-randomization variables, such as level of compliance. In particular, the complier average causal effect (CACE) - the effect of the treatment for those individuals who would comply with their treatment assignment under either treatment condition - is often of substantive interest. However, estimation of the CACE is not always straightforward, with a variety of estimation procedures and underlying assumptions, but little advice to help researchers select between methods. In this article, we discuss and examine two methods that rely on very different assumptions to estimate the CACE: a maximum likelihood ('joint') method that assumes the 'exclusion restriction,' (ER) and a propensity score-based method that relies on 'principal ignorability.' We detail the assumptions underlying each approach, and assess each methods' sensitivity to both its own assumptions and those of the other method using both simulated data and a motivating example. We find that the ER-based joint approach appears somewhat less sensitive to its assumptions, and that the performance of both methods is significantly improved when there are strong predictors of compliance. Interestingly, we also find that each method performs particularly well when the assumptions of the other approach are violated. These results highlight the importance of carefully selecting an estimation procedure whose assumptions are likely to be satisfied in practice and of having strong predictors of principal stratum membership.",
keywords = "complier average causal effect, intermediate outcomes, non-compliance, principal stratification, propensity scores",
author = "Elizabeth Stuart and Booil Jo",
year = "2015",
month = "12",
day = "1",
doi = "10.1177/0962280211421840",
language = "English (US)",
volume = "24",
pages = "657--674",
journal = "Statistical Methods in Medical Research",
issn = "0962-2802",
publisher = "SAGE Publications Ltd",
number = "6",

}

TY - JOUR

T1 - Assessing the sensitivity of methods for estimating principal causal effects

AU - Stuart, Elizabeth

AU - Jo, Booil

PY - 2015/12/1

Y1 - 2015/12/1

N2 - The framework of principal stratification provides a way to think about treatment effects conditional on post-randomization variables, such as level of compliance. In particular, the complier average causal effect (CACE) - the effect of the treatment for those individuals who would comply with their treatment assignment under either treatment condition - is often of substantive interest. However, estimation of the CACE is not always straightforward, with a variety of estimation procedures and underlying assumptions, but little advice to help researchers select between methods. In this article, we discuss and examine two methods that rely on very different assumptions to estimate the CACE: a maximum likelihood ('joint') method that assumes the 'exclusion restriction,' (ER) and a propensity score-based method that relies on 'principal ignorability.' We detail the assumptions underlying each approach, and assess each methods' sensitivity to both its own assumptions and those of the other method using both simulated data and a motivating example. We find that the ER-based joint approach appears somewhat less sensitive to its assumptions, and that the performance of both methods is significantly improved when there are strong predictors of compliance. Interestingly, we also find that each method performs particularly well when the assumptions of the other approach are violated. These results highlight the importance of carefully selecting an estimation procedure whose assumptions are likely to be satisfied in practice and of having strong predictors of principal stratum membership.

AB - The framework of principal stratification provides a way to think about treatment effects conditional on post-randomization variables, such as level of compliance. In particular, the complier average causal effect (CACE) - the effect of the treatment for those individuals who would comply with their treatment assignment under either treatment condition - is often of substantive interest. However, estimation of the CACE is not always straightforward, with a variety of estimation procedures and underlying assumptions, but little advice to help researchers select between methods. In this article, we discuss and examine two methods that rely on very different assumptions to estimate the CACE: a maximum likelihood ('joint') method that assumes the 'exclusion restriction,' (ER) and a propensity score-based method that relies on 'principal ignorability.' We detail the assumptions underlying each approach, and assess each methods' sensitivity to both its own assumptions and those of the other method using both simulated data and a motivating example. We find that the ER-based joint approach appears somewhat less sensitive to its assumptions, and that the performance of both methods is significantly improved when there are strong predictors of compliance. Interestingly, we also find that each method performs particularly well when the assumptions of the other approach are violated. These results highlight the importance of carefully selecting an estimation procedure whose assumptions are likely to be satisfied in practice and of having strong predictors of principal stratum membership.

KW - complier average causal effect

KW - intermediate outcomes

KW - non-compliance

KW - principal stratification

KW - propensity scores

UR - http://www.scopus.com/inward/record.url?scp=84948395048&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84948395048&partnerID=8YFLogxK

U2 - 10.1177/0962280211421840

DO - 10.1177/0962280211421840

M3 - Article

C2 - 21971481

AN - SCOPUS:84948395048

VL - 24

SP - 657

EP - 674

JO - Statistical Methods in Medical Research

JF - Statistical Methods in Medical Research

SN - 0962-2802

IS - 6

ER -