scholarly journals Metalearners for estimating heterogeneous treatment effects using machine learning

2019 ◽  
Vol 116 (10) ◽  
pp. 4156-4165 ◽  
Author(s):  
Sören R. Künzel ◽  
Jasjeet S. Sekhon ◽  
Peter J. Bickel ◽  
Bin Yu
Keyword(s):  
Machine Learning ◽  
Treatment Effects ◽  
Field Experiments ◽  
Average Treatment Effect ◽  
Regularity Conditions ◽  
Heterogeneous Treatment Effects ◽  
Lipschitz Continuous ◽  
Additive Regression ◽  
Underlying Mechanisms ◽  
And Control

There is growing interest in estimating and analyzing heterogeneous treatment effects in experimental and observational studies. We describe a number of metaalgorithms that can take advantage of any supervised learning or regression method in machine learning and statistics to estimate the conditional average treatment effect (CATE) function. Metaalgorithms build on base algorithms—such as random forests (RFs), Bayesian additive regression trees (BARTs), or neural networks—to estimate the CATE, a function that the base algorithms are not designed to estimate directly. We introduce a metaalgorithm, the X-learner, that is provably efficient when the number of units in one treatment group is much larger than in the other and can exploit structural properties of the CATE function. For example, if the CATE function is linear and the response functions in treatment and control are Lipschitz-continuous, the X-learner can still achieve the parametric rate under regularity conditions. We then introduce versions of the X-learner that use RF and BART as base learners. In extensive simulation studies, the X-learner performs favorably, although none of the metalearners is uniformly the best. In two persuasion field experiments from political science, we demonstrate how our X-learner can be used to target treatment regimes and to shed light on underlying mechanisms. A software package is provided that implements our methods.

scholarly journals Hybridizing Machine Learning Methods and Finite Mixture Models for Estimating Heterogeneous Treatment Effects in Latent Classes

2020 ◽  
pp. 107699862095198
Author(s):  
Youmi Suk ◽  
Jee-Seon Kim ◽  
Hyunseung Kang
Keyword(s):  
Machine Learning ◽  
Treatment Effects ◽  
Latent Class ◽  
Finite Mixture Models ◽  
Likelihood Estimation ◽  
Study Data ◽  
Latent Class Models ◽  
Finite Mixture ◽  
Latent Classes ◽  
Heterogeneous Treatment Effects

There has been increasing interest in exploring heterogeneous treatment effects using machine learning (ML) methods such as causal forests, Bayesian additive regression trees, and targeted maximum likelihood estimation. However, there is little work on applying these methods to estimate treatment effects in latent classes defined by well-established finite mixture/latent class models. This article proposes a hybrid method, a combination of finite mixture modeling and ML methods from causal inference to discover effect heterogeneity in latent classes. Our simulation study reveals that hybrid ML methods produced more precise and accurate estimates of treatment effects in latent classes. We also use hybrid ML methods to estimate the differential effects of private lessons across latent classes from Trends in International Mathematics and Science Study data.

scholarly journals Bootstrap bias correction for average treatment effects with inverse propensity weights

2019 ◽  
Vol 52 (2) ◽  
pp. 187-200
Author(s):  
GUBHINDER KUNDHI ◽  
MARCEL VOIA
Keyword(s):  
Propensity Score ◽  
Treatment Effect ◽  
Bias Correction ◽  
Treatment Effects ◽  
Potential Problem ◽  
Average Treatment Effect ◽  
Omitted Variables ◽  
Control Groups ◽  
Average Treatment ◽  
And Control

The estimated average treatment effect in observational studies is biased if the assumptions of ignorability and overlap are not satisfied. To deal with this potential problem when propensity score weights are used in the estimation of the treatment effects, in this paper we propose a bootstrap bias correction estimator for the average treatment effect (ATE) obtained with the inverse propensity score (BBC-IPS) estimator. We show in simulations that the BBC-IPC performs well when we have misspecifications of the propensity score (PS) due to: omitted variables (ignorability property may not be satisfied), overlap (imbalances in distribution between treatment and control groups) and confounding effects between observables and unobservables (endogeneity). Further refinements in bias reductions of the ATE estimates in smaller samples are attained by iterating the BBC-IPS estimator.

Broken or Fixed Effects?

2018 ◽  
Vol 8 (1) ◽  
Author(s):  
Charles E. Gibbons ◽  
Juan Carlos Suárez Serrato ◽  
Michael B. Urbancic
Keyword(s):  
Empirical Evidence ◽  
Treatment Effect ◽  
Fixed Effects ◽  
Treatment Effects ◽  
Average Treatment Effect ◽  
Significant Treatment Effect ◽  
Significant Treatment ◽  
Heterogeneous Treatment Effects ◽  
Treatment Effect Heterogeneity ◽  
Effect Heterogeneity

Abstract We replicate eight influential papers to provide empirical evidence that, in the presence of heterogeneous treatment effects, OLS with fixed effects (FE) is generally not a consistent estimator of the average treatment effect (ATE). We propose two alternative estimators that recover the ATE in the presence of group-specific heterogeneity. We document that heterogeneous treatment effects are common and the ATE is often statistically and economically different from the FE estimate. In all but one of our replications, there is statistically significant treatment effect heterogeneity and, in six, the ATEs are either economically or statistically different from the FE estimates.

scholarly journals Uncovering Individualised Treatment Effect: Evidence from Educational Trials

2020 ◽  
Author(s):  
ZhiMin Xiao ◽  
Oliver P Hauser ◽  
Charlie Kirkwood ◽  
Daniel Z. Li ◽  
Benjamin Jones ◽  
...  
Keyword(s):  
Low Income ◽  
Treatment Effect ◽  
Large Scale ◽  
Treatment Effects ◽  
Statistical Significance ◽  
Educational Interventions ◽  
Average Treatment Effect ◽  
Heterogeneous Treatment Effects ◽  
Subgroup Analyses ◽  
Individualised Treatment

The use of large-scale Randomised Controlled Trials (RCTs) is fast becoming "the gold standard" of testing the causal effects of policy, social, and educational interventions. RCTs are typically evaluated — and ultimately judged — by the economic, educational, and statistical significance of the Average Treatment Effect (ATE) in the study sample. However, many interventions have heterogeneous treatment effects across different individuals, not captured by the ATE. One way to identify heterogeneous treatment effects is to conduct subgroup analyses, such as focusing on low-income Free School Meal pupils as required for projects funded by the Education Endowment Foundation (EEF) in England. These subgroup analyses, as we demonstrate in 48 EEF-funded RCTs involving over 200,000 students, are usually not standardised across studies and offer flexible degrees of freedom to researchers, potentially leading to mixed results. Here, we develop and deploy a machine-learning and regression-based framework for systematic estimation of Individualised Treatment Effect (ITE), which can show where a seemingly ineffective and uninformative intervention worked, for whom, and by how much. Our findings have implications for decision-makers in education, public health, and medical trials.

scholarly journals Using Field Experiments to Encourage Healthy Eating in Schools

2019 ◽  
Vol 48 (3) ◽  
pp. 505-518
Author(s):  
Joseph Price
Keyword(s):  
Data Collection ◽  
Statistical Inference ◽  
Healthy Eating ◽  
Treatment Effects ◽  
Field Experiments ◽  
Heterogeneous Treatment Effects ◽  
Long Run ◽  
Practical Tool

Schools provide a unique opportunity to influence healthy eating decisions in children. Field experiments provide a practical tool for evaluating the types of interventions that can have the largest impact on these decisions in the short and long run. This article provides some insights on conducting field experiments in schools; the issues it covers are related to data collection, randomization, heterogeneous treatment effects, and statistical inference.

Sign in / Sign up

Export Citation Format

Share Document