Article
Public, Environmental & Occupational Health
Haodong Tian, Stephen Burgess
Summary: Multivariable Mendelian randomization (MVMR) is an approach that can assess the effects of related risk factors using genetic variants as instrumental variables, but performs poorly in studying time-varying causal effects, possibly due to model misspecification and violation of the exclusion restriction assumption.
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
(2023)
Article
Statistics & Probability
Andrew J. Spieker, Robert A. Greevy, Lyndsay A. Nelson, Lindsay S. Mayberry
Summary: Estimation of local average treatment effects in randomized trials relies on the exclusion restriction assumption. Recently, there has been interest in mobile health interventions, which require relaxing the exclusion restriction assumption. We propose a sensitivity analysis procedure for evaluating mobile health interventions.
ANNALS OF APPLIED STATISTICS
(2022)
Article
Economics
Martin E. Andresen, Martin Huber
Summary: This article discusses issues that arise when converting multivalued endogenous treatments to binary measures in instrumental variables estimation, and proposes assumptions and tests to address these problems.
ECONOMETRICS JOURNAL
(2021)
Article
Automation & Control Systems
Baoluo Sun, Yifan Cui, Eric Tchetgen Tchetgen
Summary: This paper discusses instrumental variable methods for identifying causal effects in the presence of unmeasured confounding. It introduces a novel approach to address the lack of credibility in the exclusion restriction assumption by proposing a multiply robust locally efficient estimator and a multiply debiased machine learning estimator. The proposed methods are evaluated through extensive simulations and a data analysis on the causal effect of 401(k) participation on savings.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Biology
Ting Ye, Ashkan Ertefaie, James Flory, Sean Hennessy, Dylan S. Small
Summary: This article introduces a causal inference method based on observational studies, which estimates treatment effects by leveraging exogenous randomness in the exposure trend. A new method called instrumented difference-in-differences is proposed, along with corresponding estimators, and their properties are analyzed. The method is also extended to a two-sample design.
Article
Statistics & Probability
Eric Tchetgen Tchetgen, BaoLuo Sun, Stefan Walter
Summary: Mendelian randomization is a popular method for recovering valid inferences about exposure-outcome causal associations using genetic markers as instrumental variables, but in practice, violation of the exclusion restriction assumption is a common concern. To address this issue, researchers have introduced a new class of IV estimators called MR GENIUS, which is robust to violation of the exclusion restriction assumption and applicable to various practical settings and causal models.
STATISTICAL SCIENCE
(2021)
Article
Statistics & Probability
Anqi Zhao, Youjin Lee, Dylan S. Small, Bikram Karmakar
Summary: This article proposes a balanced block design method to offset the possible violation of the exclusion restriction by balancing the instruments, in order to construct approximate evidence factors. It also introduces a novel stratification method for using multiple nested candidate instruments.
ANNALS OF STATISTICS
(2022)
Article
Economics
Shuo Li, Liuhua Peng, Yundong Tu
Summary: In this paper, a unified methodology is developed to test the independence assumption between the error term and exogenous variables. The methodology can be applied to a wide range of parametric models and can handle endogeneity and instrumental variables. Tests are constructed using continuous functionals and a multiple testing approach is proposed to accommodate high-dimensional exogenous random vectors. The tests are shown to be consistent and sensitive to locally alternative hypotheses.
ECONOMETRIC REVIEWS
(2022)
Article
Computer Science, Interdisciplinary Applications
Raluca Gui, Markus Meierer, Patrik Schilter, Rene Algesheimer
Summary: Endogeneity is a common issue in causal analysis when the independence assumption between an explanatory variable and the error in a statistical model is violated. Instrumental variable estimation is a possible solution, but finding valid and strong external instruments is difficult. Therefore, internal instrumental variable approaches have been proposed to correct for endogeneity without relying on external instruments. The R package REndo implements various internal instrumental variable methods.
JOURNAL OF STATISTICAL SOFTWARE
(2023)
Article
Energy & Fuels
Amaris Dalton, Bernard Bekker
Summary: This paper investigates the wind speed predictive skill provided by 16 exogenous meteorological variables in machine learning algorithms. Factors such as choice of regression model, choice of domain size, forecast period, meteorological variable elevation, and testing site location are considered. The best performing wind speed predictors were found to be 950 hPa- vertical velocity, divergence, the u- & v-wind speed components, and geopotential heights.
Article
Business, Finance
Henrique Castro Martins
Summary: This study investigates the relationship between the disclosure of financially material information and corporate risk. The findings suggest that firms with higher disclosure scores of financially material items have lower stock return standard deviation and lower idiosyncratic risk. Additionally, the research shows that the disclosure of financially material items reduces total risk but not idiosyncratic risk.
FINANCE RESEARCH LETTERS
(2023)
Article
Statistics & Probability
Gonzalo Vazquez-Bare
Summary: This study proposes a potential outcomes framework to analyze spillover effects using instrumental variables. The findings suggest that intention-to-treat (ITT) parameters do not have a clear link to causally interpretable parameters, and rescaling them by first-stage estimands recovers a weighted combination of average effects. The study also analyzes the identification of causal effects under one-sided noncompliance and introduces an alternative assumption for identifying parameters of interest under two-sided noncompliance.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2023)
Article
Statistics & Probability
Frank Windmeijer, Xiaoran Liang, Fernando P. Hartwig, Jack Bowden
Summary: The CI method proposed in this study is based on confidence intervals to select valid instruments from a larger set, ensuring a monotonically decreasing number of instruments selected with decreasing tuning parameter values. The method employs a downward testing procedure with the Sargan test as a major verification step.
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES B-STATISTICAL METHODOLOGY
(2021)
Article
Economics
Juan Carlos Escanciano, Wei Li
Summary: This paper studies the identification and estimation of the optimal linear approximation of a structural regression function, introduces a Two-Step IV estimator based on Tikhonov regularization, and verifies its asymptotic normality without completeness or identification. Monte Carlo simulations suggest excellent finite sample performance for the proposed inferences.
JOURNAL OF ECONOMETRICS
(2021)
Article
Computer Science, Information Systems
Chengxiang Dong, Xiaoliang Feng, Yongchao Wang, Xin Wei
Summary: Traffic flow prediction is crucial for Intelligent Transportation Systems (ITS), but accurately predicting traffic flow for a large-scale road network is challenging due to complex and dynamic spatiotemporal dependencies. In this paper, we propose a Spatiotemporal Exogenous Variables Enhanced Transformer (SEE-Transformer) model that incorporates Graph Attention Networks and Transformer architectures to capture these dependencies. The model leverages rich exogenous variables and constructs traffic graphs based on sensor connections and pattern similarity for improved prediction accuracy.