Reducing Model Misspecification and Bias in the Estimation of Interactions
Bibliographic Data
| ID | 6331978 |
|---|---|
| Authors | Matthew Blackwell (0000-0002-3689-9527, Harvard University, corresponding author), Michael P Olson (0000-0002-8023-1815, Washington University in St. Louis) |
| Year | 2022 |
| Volume | 30 |
| Issue | 4 |
| Pages | 495-514 |
| Publication date | 2022-10-01 |
| Peer Reviewed | Yes |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | Political Analysis (JOURNAL) |
| Journal identifiers | ISSN: 1047-1987 • E-ISSN: 1476-4989 |
| Publisher | Cambridge University Press (CUP) (PUBLISHER) |
| DOI | 10.1017/pan.2021.19 |
| OpenAlex | W3184944157 |
| Language | EN |
| Citations received | 38 |
| References cited | 29 |
Analyzing variation in treatment effects across subsets of the population is an important way for social scientists to evaluate theoretical arguments. A common strategy in assessing such treatment effect heterogeneity is to include a multiplicative interaction term between the treatment and a hypothesized effect modifier in a regression model. Unfortunately, this approach can result in biased inferences due to unmodeled interactions between the effect modifier and other covariates, and including these interactions can lead to unstable estimates due to overfitting. In this paper, we explore the usefulness of machine learning algorithms for stabilizing these estimates and show how many off-the-shelf adaptive methods lead to two forms of bias: direct and indirect regularization bias. To overcome these issues, we use a post-double selection approach that utilizes several lasso estimators to select the interactions to include in the final model. We extend this approach to estimate uncertainty for both interaction and marginal effects. Simulation evidence shows that this approach has better performance than competing methods, even when the number of covariates is large. We show in two empirical examples that the choice of method leads to dramatically different conclusions about effect heterogeneity
Artificial neural network · Covariate · Econometrics · Estimator · Interaction · Lasso (programming language · Machine learning · Model selection · Multiplicative function · Overfitting · Regression · Regularization (linguistics · Statistics · Advanced Causal Inference Techniques · Computer Science · Mathematics · Statistical Methods and Bayesian Inference · Statistical Methods and Inference · Artificial Intelligence
Occupation-Specific Skills, Labor Market Context, and Preferences for Redistribution
Reactive Cosmopolitanism
Shouldering childhood
Quantities of Interest for Interactions and the Pitfalls of Assuming Linear Treatment Effects
The Dual Impact of Organic Farming Support
Occupational labor markets, skills, and immigration concerns
A representative survey experiment of motivated climate change denial
Legislative Organization and Political Representation
Causal Machine Learning for Moderation Effects
Impacts of Social Assistance During the Covid‐19 Pandemic in Poor Remote Communities
Gender Gaps in Immigrants' Political Participation within and across Borders
Inequality and Total Effect Summary Measures for Nominal and Ordinal Variables
Framing electricity plan choices to enhance green energy usage
Generalizing Trimming Bounds for Endogenously Missing Outcome Data Using Random Forests
Ethnic politics and sovereign credit risk
Automated decision-making in public administration
How information, social norms, and experience with novel meat substitutes can create positive political feedback and demand-side policy change
Geography, development, and power
Race, legislative speech, and symbolic representation in Congress
Do Political Preferences Affect Policy Learning and Uptake? Evidence from a Field Experiment with Informal Entrepreneurs
Americans’ Responses to Covid-19 and the Conditional Role of Dispositional Needs for Security
The Heterogeneous Associations of Rural Consciousness and Political Preferences
Capturing the Fourth Estate
Incorporating Machine Learning into Sociological Model-Building
The Consequences of Model Misspecification for the Estimation of Nonlinear Interaction Effects
Longitudinal attenuation in political polarization
Group Ties amid Industrial Change
Sources of Partisan Change
Household education gaps and gender role attitudes
Technology Threats to Employment, Issues, and Candidate and Party Preferences in the United States
Affect, Not Ideology
Reducing Model Misspecification and Bias in the Estimation of Interactions
Carbon inequality and support for carbon taxation
Informed or Overwhelmed? Disentangling the Effects of Cognitive Ability and Information on Public Opinion
Why Anxious People Lean to the Left on Economic Policy
When expectations backfire
Relationship between English proficiency and socioeconomic status in Asia
Does taking parental leave make you a more likeable worker? Evidence from a survey experiment
The American Direct Primary
Bart
The Adaptive Lasso and Its Oracle Properties
Some heteroskedasticity-consistent covariance matrix estimators with improved finite sample properties
Inference on Treatment Effects after Selection among High-Dimensional Controls
Estimating treatment effect heterogeneity in randomized program evaluation
Regression Shrinkage and Selection Via the Lasso
Inference in High-Dimensional Panel Models With an Application to Gun Control
Nonparametric Estimation of Average Treatment Effects Under Exogeneity
The Decline of Third-Party Voting in the United States
Problems with products? Control strategies for models with interaction and quadratic effects
The causal interpretation of estimated associations in regression models
At the Nexus of Observational and Experimental Research
Reducing Model Misspecification and Bias in the Estimation of Interactions
Sparse Estimation and Uncertainty with Application to Subgroup Analysis
Understanding Interaction Models
How Much Should We Trust Estimates from Multiplicative Interaction Models? Simple Tools to Improve Empirical Practice
Kernel Regularized Least Squares
Hypothesis Testing and Multiplicative Interaction Terms
Marginal Effects in Interaction Models
Remittances and Protest in Dictatorships
Testing for Interaction in Binary Logit and Probit Models
Modeling and Interpreting Interactive Hypotheses in Regression Analysis
| Unique citing works | 38 |
|---|---|
| Citations per year | 6,33 |
| Citation span | 2020 - 2026 (7) |
| Citation velocity | current |
| Highly cited | No |
| Citation types | Neutral: 37 |