scholarly journals Robust Machine Learning for Treatment Effects in Multilevel Observational Studies Under Cluster-level Unmeasured Confounding

2020 ◽  
Author(s):  
Youmi Suk ◽  
Hyunseung Kang
Keyword(s):  
Machine Learning ◽  
Observational Studies ◽  
Treatment Effects ◽  
Math Achievement ◽  
R Package ◽  
Simulation Studies ◽  
Unmeasured Confounding ◽  
Doubly Robust ◽  
Unmeasured Confounders ◽  
Cluster Level

Recently, machine learning (ML) methods have been used in causal inference to estimate treatment effects in order to reduce concerns for model mis-specification. However, many, if not all, ML methods require that all confounders are measured to consistently estimate treatment effects. In this paper, we propose a family of ML methods that estimate treatment effects in the presence of cluster-level unmeasured confounders, a type of unmeasured confounders that are shared within each cluster and are common in multilevel observational studies. We show through simulation studies that our proposed methods are consistent and doubly robust when unmeasured cluster-level confounders are present. We also examine the effect of taking an algebra course on math achievement scores from the Early Childhood Longitudinal Study, a multilevel observational educational study, using our methods. The proposed methods are available in the CURobustML R package.

scholarly journals 1455Leveraging multi-omic negative controls for effect estimation in molecular epidemiologic studies: A simulation study

2021 ◽  
Vol 50 (Supplement_1) ◽  
Author(s):  
Jonathan Huang
Keyword(s):  
Machine Learning ◽  
Molecular Epidemiology ◽  
Maternal Blood ◽  
Epidemiologic Studies ◽  
Negative Control ◽  
Biological Knowledge ◽  
Methylation Array ◽  
Unmeasured Confounding ◽  
Dna Methylation Array ◽  
Doubly Robust

Abstract Background Exploratory null-hypothesis significance testing (e.g. GWAS, EWAS) form the backbone of molecular epidemiology, however methods to identify true causal signals are underdeveloped. Via plasmode simulation, I evaluate two approaches to quantitatively control for shared unmeasured confounding and recover unbiased effects using complementary epigenomes and biologically-informed structural assumptions. Methods I adapt proposed negative control-based estimators, the control outcome calibration approach (COCA) and proximal g-computation (PG) to case studies in perinatal molecular epidemiology. COCA may be employed when maternal epigenome has no direct effects on phenotype and proxy shared unmeasured confounders and PG further with suitable genetic instruments (e.g. mQTLs). Baseline covariates were extracted from 777 mother-child pairs in a birth cohort with maternal blood and fetal cord DNA methylation array data. Treatment and outcome values were simulated in 2000 bootstraps. Bootstrapped, ordinary (COCA) and 2-stage (PG) least squares were fitted to estimate treatment effects and standard errors under various common settings of missing confounders (e.g. paternal data). Doubly-robust, machine learning estimators were explored. Results COCA and PG performed well in simplistic data generating processes. However, in real-world cohort simulations, COCA performed acceptably only in settings with strong proxy confounders, but otherwise poorly (median bias 610%; coverage 29%). PG performed slightly better. Alternatively, simple covariate adjustment for maternal methylation outperformed (median bias 22%; 71% coverage) COCA, PG, and machine learning estimators. Discussion Molecular epidemiology provides key opportunity to leverage biological knowledge against unmeasured confounding. Negative control calibration or adjustments may help under limited scenarios where assumptions are fulfilled, but should be tested with suitable simulations. Key messages Quantitative approaches for unmeasured confounding in molecular epidemiology are a critical gap. Negative control calibration or adjustment may help under limiting scenarios. Proposed estimators should be tested in simulation settings that closely mimic target data.

Combining planned and discovered comparisons in observational studies

Biostatistics ◽  
2018 ◽  
Vol 21 (3) ◽  
pp. 384-399 ◽  
Author(s):  
Paul R Rosenbaum
Keyword(s):  
Cognitive Decline ◽  
Observational Studies ◽  
Multiple Testing ◽  
Joint Distribution ◽  
Treatment Effects ◽  
A Priori ◽  
Principal Component ◽  
R Package ◽  
Test Statistic

Summary In observational studies of treatment effects, it is common to have several outcomes, perhaps of uncertain quality and relevance, each purporting to measure the effect of the treatment. A single planned combination of several outcomes may increase both power and insensitivity to unmeasured bias when the plan is wisely chosen, but it may miss opportunities in other cases. A method is proposed that uses one planned combination with only a mild correction for multiple testing and exhaustive consideration of all possible combinations fully correcting for multiple testing. The method works with the joint distribution of $\kappa^{T}\left( \mathbf{T}-\boldsymbol{\mu}\right) /\sqrt {\boldsymbol{\kappa}^{T}\boldsymbol{\Sigma\boldsymbol{\kappa}}}$ and $max_{\boldsymbol{\lambda}\neq\mathbf{0}}$$\,\lambda^{T}\left( \mathbf{T} -\boldsymbol{\mu}\right) /$$\sqrt{\boldsymbol{\lambda}^{T}\boldsymbol{\Sigma \lambda}}$ where $\kappa$ is chosen a priori and the test statistic $\mathbf{T}$ is asymptotically $N_{L}\left( \boldsymbol{\mu},\boldsymbol{\Sigma}\right) $. The correction for multiple testing has a smaller effect on the power of $\kappa^{T}\left( \mathbf{T}-\boldsymbol{\mu }\right) /\sqrt{\boldsymbol{\kappa}^{T}\boldsymbol{\Sigma\boldsymbol{\kappa} }}$ than does switching to a two-tailed test, even though the opposite tail does receive consideration when $\lambda=-\kappa$. In the application, there are three measures of cognitive decline, and the a priori comparison $\kappa$ is their first principal component, computed without reference to treatment assignments. The method is implemented in an R package sensitivitymult.

scholarly journals Assessing The Effectiveness of Empirical Calibration Under Different Bias Scenarios

2021 ◽  
Author(s):  
Hon Hwang ◽  
Juan C Quiroz ◽  
Blanca Gallego
Keyword(s):  
Confidence Interval ◽  
Confidence Intervals ◽  
Observational Studies ◽  
Treatment Effects ◽  
Model Misspecification ◽  
Calibration Procedure ◽  
Negative Control ◽  
Unmeasured Confounding ◽  
Empirical Calibration ◽  
Negative Controls

Abstract Background: Estimations of causal effects from observational data are subject to various sources of bias. These biases can be adjusted by using negative control outcomes not affected by the treatment. The empirical calibration procedure uses negative controls to calibrate p-values and both negative and positive controls to calibrate coverage of the 95% confidence interval of the outcome of interest. Although empirical calibration has been used in several large observational studies, there is no systematic examination of its effect under different bias scenarios. Methods: The effect of empirical calibration of confidence intervals was analyzed using simulated datasets with known treatment effects. The simulations were for binary treatment and binary outcome, with simulated biases resulting from unmeasured confounder, model misspecification, measurement error, and lack of positivity. The performance of empirical calibration was evaluated by determining the change of the confidence interval coverage and bias of the outcome of interest. Results: Empirical calibration increased coverage of the outcome of interest by the 95% confidence interval under most settings but was inconsistent in adjusting the bias of the outcome of interest. Empirical calibration was most effective when adjusting for unmeasured confounding bias. Suitable negative controls had a large impact on the adjustment made by empirical calibration, but small improvements in the coverage of the outcome of interest was also observable when using unsuitable negative controls. Conclusions: This work adds evidence to the efficacy of empirical calibration on calibrating the confidence intervals of treatment effects in observational studies. We recommend empirical calibration of confidence intervals, especially when there is a risk of unmeasured confounding.

A conditional test with demonstrated insensitivity to unmeasured bias in matched observational studies

Biometrika ◽  
2020 ◽  
Vol 107 (4) ◽  
pp. 827-840
Author(s):  
P R Rosenbaum
Keyword(s):  
Observational Study ◽  
Null Hypothesis ◽  
Observational Studies ◽  
Treatment Effects ◽  
R Package ◽  
Causal Effects ◽  
Rank Statistic ◽  
Matched Pairs ◽  
Response Patterns ◽  
Signed Rank

Summary In an observational study matched for observed covariates, an association between treatment received and outcome exhibited may indicate not an effect caused by the treatment, but merely some bias in the allocation of treatments to individuals within matched pairs. The evidence that distinguishes moderate biases from causal effects is unevenly dispersed among possible comparisons in an observational study: some comparisons are insensitive to larger biases than others. Intuitively, larger treatment effects tend to be insensitive to larger unmeasured biases, and perhaps matched pairs can be grouped using covariates, doses or response patterns so that groups of pairs with larger treatment effects may be identified. Even if an investigator has a reasoned conjecture about where to look for insensitive comparisons, that conjecture might prove mistaken, or, when not mistaken, it might be received sceptically by other scientists who doubt the conjecture or judge it to be too convenient in light of its success with the data at hand. In this article a test is proposed that searches for insensitive findings over many comparisons, but controls the probability of falsely rejecting a true null hypothesis of no treatment effect in the presence of a bias of specified magnitude. An example is studied in which the test considers many comparisons and locates an interpretable comparison that is insensitive to larger biases than a conventional comparison based on Wilcoxon’s signed rank statistic applied to all pairs. A simulation examines the power of the proposed test. The method is implemented in the R package dstat, which contains the example and reproduces the analysis.

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 The Impact of Unmeasured Confounding in Observational Studies: a Plasmode Simulation Study of Targeted Maximum Likelihood Estimation

2021 ◽  
Author(s):  
Lateef Amusa ◽  
Temesgen Zewotir ◽  
Delia North
Keyword(s):  
Maximum Likelihood ◽  
Maximum Likelihood Estimation ◽  
Simulation Study ◽  
Observational Studies ◽  
Likelihood Estimation ◽  
Unmeasured Confounding ◽  
Targeted Maximum Likelihood Estimation ◽  
Targeted Maximum Likelihood ◽  
The Impact ◽  
Unmeasured Confounders

Abstract Unmeasured confounding can cause considerable problems in observational studies and may threaten the validity of the estimates of causal treatment effects. There has been discussion on the amount of bias in treatment effect estimates that can occur due to unmeasured confounding. We investigate the robustness of a relatively new causal inference technique, targeted maximum likelihood estimation (TMLE), in terms of its robustness to the impact of unmeasured confounders. We benchmark TMLE’s performance with the inverse probability of treatment weighting (IPW) method. We utilize a plasmode-like simulation based on variables and parameters from the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT). We evaluated the accuracy and precision of the estimated treatment effects. Though TMLE performed better in most of the scenarios considered, our simulation study results suggest that both methods performed reasonably well in estimating the marginal odds ratio, in the presence of unmeasured confounding. Nonetheless, the only remedy to unobserved confounding is controlling for as many as available covariates in an observational study, because not even TMLE can provide safeguard against bias from unmeasured confounders.

scholarly journals Sensitivity analysis for unmeasured confounding in meta-analyses

2017 ◽  
Author(s):  
Maya B Mathur ◽  
Tyler VanderWeele
Keyword(s):  
Random Effects ◽  
Observational Studies ◽  
R Package ◽  
Sensitivity Analyses ◽  
Unmeasured Confounding ◽  
Confounding Bias ◽  
Point Estimates ◽  
Bias Factor ◽  
Meta Analyses ◽  
Biased Estimates

Random-effects meta-analyses of observational studies can produce biased estimates if the synthesized studies are subject to unmeasured confounding. We propose sensitivity analyses quantifying the extent to which unmeasured confounding of specified magnitude could reduce to below a certain threshold the proportion of true effect sizes that are scientifically meaningful. We also develop converse methods to estimate the strength of confounding capable of reducing the proportion of scientifically meaningful true effects to below a chosen threshold. These methods apply when a "bias factor'' is assumed to be normally distributed across studies or is assessed across a range of fixed values. Our estimators are derived using recently proposed sharp bounds on confounding bias within a single study that do not make assumptions regarding the unmeasured confounders themselves or the functional form of their relationships to the exposure and outcome of interest. We provide an R package, EValue, and a free website that compute point estimates and inference and produce plots for conducting such sensitivity analyses. These methods facilitate principled use of random-effects meta-analyses of observational studies to assess the strength of causal evidence for a hypothesis.

EpiPen: An R Package to Investigate Two-Locus Epistatic Models

2014 ◽  
Vol 17 (4) ◽  
Author(s):  
Raymond K. Walters ◽  
Charles Laurin ◽  
Gitta H. Lubke
Keyword(s):  
Power Analysis ◽  
R Package ◽  
Simulation Studies ◽  
Nucleotide Polymorphisms ◽  
Single Nucleotide ◽  
Epistatic Interactions ◽  
Model Interpretation ◽  
Genome Wide ◽  
Using Data ◽  
Power Analyses

Epistasis is a growing area of research in genome-wide studies, but the differences between alternative definitions of epistasis remain a source of confusion for many researchers. One problem is that models for epistasis are presented in a number of formats, some of which have difficult-to-interpret parameters. In addition, the relation between the different models is rarely explained. Existing software for testing epistatic interactions between single-nucleotide polymorphisms (SNPs) does not provide the flexibility to compare the available model parameterizations. For that reason we have developed an R package for investigating epistatic and penetrance models, EpiPen, to aid users who wish to easily compare, interpret, and utilize models for two-locus epistatic interactions. EpiPen facilitates research on SNP-SNP interactions by allowing the R user to easily convert between common parametric forms for two-locus interactions, generate data for simulation studies, and perform power analyses for the selected model with a continuous or dichotomous phenotype. The usefulness of the package for model interpretation and power analysis is illustrated using data on rheumatoid arthritis.

scholarly journals stochprofML: stochastic profiling using maximum likelihood estimation in R

2021 ◽  
Vol 22 (1) ◽  
Author(s):  
Lisa Amrhein ◽  
Christiane Fuchs
Keyword(s):  
Maximum Likelihood ◽  
Cell Fate ◽  
Likelihood Estimation ◽  
R Package ◽  
Cell Populations ◽  
Simulation Studies ◽  
Likelihood Principle ◽  
Maximum Likelihood Principle ◽  
Molecular Expression ◽  
Mixed Samples

Abstract Background Tissues are often heterogeneous in their single-cell molecular expression, and this can govern the regulation of cell fate. For the understanding of development and disease, it is important to quantify heterogeneity in a given tissue. Results We present the R package stochprofML which uses the maximum likelihood principle to parameterize heterogeneity from the cumulative expression of small random pools of cells. We evaluate the algorithm’s performance in simulation studies and present further application opportunities. Conclusion Stochastic profiling outweighs the necessary demixing of mixed samples with a saving in experimental cost and effort and less measurement error. It offers possibilities for parameterizing heterogeneity, estimating underlying pool compositions and detecting differences between cell populations between samples.

Sign in / Sign up

Export Citation Format

Share Document