Unifying instrumental variable and inverse probability weighting approaches for inference of causal treatment effect and unmeasured confounding in observational studies

2020 ◽  
pp. 096228022097183
Author(s):  
Tao Liu ◽  
Joseph W Hogan
Keyword(s):  
Treatment Effect ◽  
Instrumental Variable ◽  
Observational Studies ◽  
Treatment Effects ◽  
Inverse Probability Weighting ◽  
Probability Weighting ◽  
Unmeasured Confounding ◽  
Inverse Probability ◽  
Causal Treatment ◽  
Weighting Methods

Confounding is a major concern when using data from observational studies to infer the causal effect of a treatment. Instrumental variables, when available, have been used to construct bound estimates on population average treatment effects when outcomes are binary and unmeasured confounding exists. With continuous outcomes, meaningful bounds are more challenging to obtain because the domain of the outcome is unrestricted. In this paper, we propose to unify the instrumental variable and inverse probability weighting methods, together with suitable assumptions in the context of an observational study, to construct meaningful bounds on causal treatment effects. The contextual assumptions are imposed in terms of the potential outcomes that are partially identified by data. The inverse probability weighting component incorporates a sensitivity parameter to encode the effect of unmeasured confounding. The instrumental variable and inverse probability weighting methods are unified using the principal stratification. By solving the resulting system of estimating equations, we are able to quantify both the causal treatment effect and the sensitivity parameter (i.e. the degree of the unmeasured confounding). We demonstrate our method by analyzing data from the HIV Epidemiology Research Study.

scholarly journals Statistical considerations in the design and analysis of non-inferiority trials with binary endpoints in the presence of non-adherence: a simulation study

2020 ◽  
Vol 4 ◽  
pp. 207 ◽  
Author(s):  
Yin Mo ◽  
Cherry Lim ◽  
Mavuto Mukaka ◽  
Ben S. Cooper
Keyword(s):  
Treatment Effect ◽  
Instrumental Variable ◽  
Experimental Treatment ◽  
Inverse Probability Weighting ◽  
Alternative Methods ◽  
Sensitivity Analyses ◽  
Probability Weighting ◽  
Inverse Probability ◽  
Instrumental Variable Estimation ◽  
The Impact

Protocol non-adherence is common and poses unique challenges in the interpretation of trial outcomes, especially in non-inferiority trials. We performed simulations of a non-inferiority trial with a time-fixed treatment and a binary endpoint in order to: i) explore the impact of various patterns of non-adherence and analysis methods on treatment effect estimates; ii) quantify the probability of claiming non-inferiority when the experimental treatment effect is actually inferior; and iii) evaluate alternative methods such as inverse probability weighting and instrumental variable estimation. We found that the probability of concluding non-inferiority when the experimental treatment is actually inferior depends on whether non-adherence is due to confounding or non-confounding factors, and the actual treatments received by the non-adherent participants. With non-adherence, intention-to-treat analysis has a higher tendency to conclude non-inferiority when the experimental treatment is actually inferior under most patterns of non-adherence. This probability of concluding non-inferiority can be increased to as high as 0.1 from 0.025 when the adherence is relatively high at 90%. The direction of bias for the per-protocol analysis depends on the directions of influence the confounders have on adherence and probability of outcome. The inverse probability weighting approach can reduce bias but will only eliminate it if all confounders can be measured without error and are appropriately adjusted for. Instrumental variable estimation overcomes this limitation and gives unbiased estimates even when confounders are not known, but typically requires large sample sizes to achieve acceptable power. Investigators need to consider patterns of non-adherence and potential confounders in trial designs. Adjusted analysis of the per-protocol population with sensitivity analyses on confounders and other approaches, such as instrumental variable estimation, should be considered when non-compliance is anticipated. We provide an online power calculator allowing for various patterns of non-adherence using the above methods.

scholarly journals Statistical considerations in the design and analysis of non-inferiority trials with binary endpoints in the presence of non-adherence: a simulation study

2019 ◽  
Vol 4 ◽  
pp. 207
Author(s):  
Yin Mo ◽  
Cherry Lim ◽  
Mavuto Mukaka ◽  
Ben S. Cooper
Keyword(s):  
Treatment Effect ◽  
Instrumental Variable ◽  
Experimental Treatment ◽  
Inverse Probability Weighting ◽  
Alternative Methods ◽  
Sensitivity Analyses ◽  
Probability Weighting ◽  
Inverse Probability ◽  
Instrumental Variable Estimation ◽  
The Impact

Protocol non-adherence is common and poses unique challenges in the interpretation of trial outcomes, especially in non-inferiority trials. We performed simulations of a non-inferiority trial with a time-fixed treatment and a binary endpoint in order to: i) explore the impact of various patterns of non-adherence and analysis methods on treatment effect estimates; ii) quantify the probability of claiming non-inferiority when the experimental treatment effect is actually inferior; and iii) evaluate alternative methods such as inverse probability weighting and instrumental variable estimation. We found that the probability of concluding non-inferiority when the experimental treatment is actually inferior depends on whether non-adherence is due to confounding or non-confounding factors, and the actual treatments received by the non-adherent participants. With non-adherence, intention-to-treat analysis has a higher tendency to conclude non-inferiority when the experimental treatment is actually inferior under most patterns of non-adherence. This probability of concluding non-inferiority can be increased to as high as 0.1 from 0.025 when the adherence is relatively high at 90%. The direction of bias for the per-protocol analysis depends on the directions of influence the confounders have on adherence and probability of outcome. The inverse probability weighting approach can reduce bias but will only eliminate it if all confounders can be measured without error and are appropriately adjusted for. Instrumental variable estimation overcomes this limitation and gives unbiased estimates even when confounders are not known, but typically requires large sample sizes to achieve acceptable power. Investigators need to consider patterns of non-adherence and potential confounders in trial designs. Adjusted analysis of the per-protocol population with sensitivity analyses on confounders and other approaches, such as instrumental variable estimation, should be considered when non-compliance is anticipated. We provide an online power calculator allowing for various patterns of non-adherence using the above methods.

Estimating Treatment Effects in the Presence of Noncompliance and Nonresponse: The Generalized Endogenous Treatment Model

2011 ◽  
Vol 19 (2) ◽  
pp. 205-226 ◽  
Author(s):  
Kevin M. Esterling ◽  
Michael A. Neblo ◽  
David M. J. Lazer
Keyword(s):  
Treatment Effect ◽  
Instrumental Variable ◽  
Treatment Effects ◽  
Mean Squared Error ◽  
Principal Stratification ◽  
Squared Error ◽  
Treatment Effect Heterogeneity ◽  
Causal Treatment ◽  
Effect Heterogeneity ◽  
And Control

If ignored, noncompliance with a treatment or nonresponse on outcome measures can bias estimates of treatment effects in a randomized experiment. To identify and estimate causal treatment effects in the case where compliance and response depend on unobservables, we propose the parametric generalized endogenous treatment (GET) model. GET incorporates behavioral responses within an experiment to measure each subject's latent compliance type and identifies causal effects via principal stratification. Using simulation methods and an application to field experimental data, we show GET has a dramatically lower mean squared error for treatment effect estimates than existing approaches to principal stratification that impute, rather than measure, compliance type. In addition, we show that GET allows one to relax and test the instrumental variable exclusion restriction assumption, to test for the presence of treatment effect heterogeneity across a range of compliance types, and to test for treatment ignorability when treatment and control samples are balanced on observable covariates.

Comparing Performance of Methods to Deal With Differential Attrition in Randomized Experimental Evaluations

2021 ◽  
Vol 45 (1-2) ◽  
pp. 70-104
Author(s):  
Kaitlin Anderson ◽  
Gema Zamarro ◽  
Jennifer Steele ◽  
Trey Miller
Keyword(s):  
Randomized Controlled Trials ◽  
Research Design ◽  
Treatment Effect ◽  
Controlled Trials ◽  
Probability Weighting ◽  
Attrition Rates ◽  
Inverse Probability ◽  
Randomized Controlled ◽  
Two Samples ◽  
Weighting Methods

Background: In randomized controlled trials, attrition rates often differ by treatment status, jeopardizing causal inference. Inverse probability weighting methods and estimation of treatment effect bounds have been used to adjust for this bias. Objectives: We compare the performance of various methods within two samples, both generated through lottery-based randomization: one with considerable differential attrition and an augmented dataset with less problematic attrition. Research Design: We assess the performance of various correction methods within the dataset with problematic attrition. In addition, we conduct simulation analyses. Results: Within the more problematic dataset, we find the correction methods often performed poorly. Simulation analyses indicate that deviations from the underlying assumptions for bounding approaches damage the performance of estimated bounds. Conclusions: We recommend the verification of the underlying assumptions in attrition correction methods whenever possible and, when verification is not possible, using these methods with caution.

Causal inference with measurement error in outcomes: Bias analysis and estimation methods

2017 ◽  
Vol 28 (7) ◽  
pp. 2049-2068 ◽  
Author(s):  
Di Shu ◽  
Grace Y Yi
Keyword(s):  
Measurement Error ◽  
Treatment Effect ◽  
Inverse Probability Weighting ◽  
Average Treatment Effect ◽  
Estimation Methods ◽  
Outcome Variable ◽  
Probability Weighting ◽  
Validation Data ◽  
Inverse Probability ◽  
Average Treatment

Inverse probability weighting estimation has been popularly used to consistently estimate the average treatment effect. Its validity, however, is challenged by the presence of error-prone variables. In this paper, we explore the inverse probability weighting estimation with mismeasured outcome variables. We study the impact of measurement error for both continuous and discrete outcome variables and reveal interesting consequences of the naive analysis which ignores measurement error. When a continuous outcome variable is mismeasured under an additive measurement error model, the naive analysis may still yield a consistent estimator; when the outcome is binary, we derive the asymptotic bias in a closed-form. Furthermore, we develop consistent estimation procedures for practical scenarios where either validation data or replicates are available. With validation data, we propose an efficient method for estimation of average treatment effect; the efficiency gain is substantial relative to usual methods of using validation data. To provide protection against model misspecification, we further propose a doubly robust estimator which is consistent even when either the treatment model or the outcome model is misspecified. Simulation studies are reported to assess the performance of the proposed methods. An application to a smoking cessation dataset is presented.

Sign in / Sign up

Export Citation Format

Share Document