A5049923911- Statistical Methods and Inference
- Advanced Causal Inference Techniques
- Statistical Methods and Bayesian Inference
- Electoral Systems and Political Participation
- Health Systems, Economic Evaluations, Quality of Life
- Advanced Statistical Methods and Models
- Healthcare Policy and Management
- School Choice and Performance
- Statistical Methods in Clinical Trials
- Bayesian Modeling and Causal Inference
- Survey Methodology and Nonresponse
- Experimental Behavioral Economics Studies
- Random Matrices and Applications
- Financial Risk and Volatility Modeling
- Game Theory and Voting Systems
- Machine Learning in Healthcare
- Primary Care and Health Outcomes
- Advanced Clustering Algorithms Research
- Vaccine Coverage and Hesitancy
- Social and Intergroup Psychology
- Social Media and Politics
- Qualitative Comparative Analysis Research
- Gender Politics and Representation
- Political Conflict and Governance
- Limits and Structures in Graph Theory
Yale University
2020-2024
University of California, Berkeley
2011-2023
University of Southern California
2023
Berkeley College
2008-2020
Rutgers, The State University of New Jersey
2020
Rutgers Sexual and Reproductive Health and Rights
2020
Carnegie Mellon University
2020
Harvard University
2002-2019
University of Michigan
2017
Fortis Healthcare
2015
<b>Matching</b> is an R package which provides functions for multivariate and propensity score matching finding optimal covariate balance based on a genetic search algorithm. A variety of univariate metrics to determine if actually has been obtained are provided. The underlying algorithm written in C++, makes extensive use system BLAS scales efficiently with dataset size. finds parallelized can make multiple CPUs or cluster computers. large number options provided control exactly how the...
This paper presents genetic matching, a method of multivariate matching that uses an evolutionary search algorithm to determine the weight each covariate is given. Both propensity score and based on Mahalanobis distance are limiting cases this method. The makes transparent certain issues all methods must confront. We present simulation studies show improves balance it may reduce bias if selection observables assumption holds. then reanalysis number data sets in LaLonde (1986) controversy.
XTRACORPOREAL MEMBRANE OXYgenation (ECMO) can support gas exchange independently of mechanical ventilation in patients with severe acute respiratory failure.ECMO may be used either as a rescue intervention or to minimize ventilator-associated lung injury 1 and its associated multiple organ dysfunction, 2 both crucial determinants survival for distress syndrome (ARDS).
Significance Estimating and analyzing heterogeneous treatment effects is timely, yet challenging. We introduce a unifying framework for many conditional average effect estimators, we propose metalearner, the X-learner, which can adapt to structural properties, such as smoothness sparsity of underlying effect. present its favorable using theory simulations. apply it, random forests, two field experiments in political science, where it shown be easy use produce results that are interpretable.
<b>genoud</b> is an R function that combines evolutionary algorithm methods with a derivative-based (quasi-Newton) method to solve difficult optimization problems. may also be used for problems which derivatives do not exist. solves are nonlinear or perhaps even discontinuous in the parameters of optimized. When optimized (for example, log-likelihood) model's parameters, will generally globally concave and have irregularities such as saddlepoints discontinuities. Optimization rely on...
Following David Lee's pioneering work, numerous scholars have applied the regression discontinuity (RD) design to popular elections. Contrary assumptions of RD, however, we show that bare winners and losers in U.S. House elections (1942–2008) differ markedly on pretreatment covariates. Bare possess large ex ante financial, experience, incumbency advantages over their opponents are usually candidates predicted win by Congressional Quarterly 's pre-election ratings. Covariate imbalance...
In recent years, there has been a burst of innovative work on methods for estimating causal effects using observational data. Much this extended and brought renewed focus old approaches such as matching, which is the review. The new developments highlight an tension in social sciences: research design versus quantitative models. This realization, along with interest field experiments, marked return foundational questions opposed to fascination latest estimator. I use studies get-out-the-vote...
Difference-in-differences (DiD) estimators provide unbiased treatment effect estimates when, in the absence of treatment, average outcomes for treated and control groups would have followed parallel trends over time. This assumption is implausible many settings. An alternative that potential are independent status, conditional on past outcomes. paper considers three methods share this assumption: synthetic method, a lagged dependent variable (LDV) regression approach, matching Our motivating...
Headline Evaluating service innovations in health care and public requires flexibility, collaboration pragmatism; this collection identifies robust, innovative mixed methods to inform such evaluations.
We show that the butterfly ballot used in Palm Beach County, Florida, 2000 presidential election caused more than 2,000 Democratic voters to vote by mistake for Reform candidate Pat Buchanan, a number larger George W. Bush's certified margin of victory Florida. use multiple methods and several kinds data rule out alternative explanations votes Buchanan received County. Among 3,053 U.S. counties where was on ballot, County has most anomalous excess him. In Buchanan's proportion election-day...
Natural experiments help to overcome some of the obstacles researchers face when making causal inferences in social sciences. However, even natural interventions are randomly assigned, treatment–control comparisons made available by may not be valid. We offer a framework for clarifying issues involved, which subtle and often overlooked. illustrate our examining four different used literature. In each case, random assignment intervention is sufficient provide an unbiased estimate effect....
We look at conventional methods for removing endogeneity bias in regression models, including the linear model and probit model. It is known that usual Heckman two-step procedure should not be used model: from a theoretical perspective, it unsatisfactory, likelihood are superior. However, serious numerical problems occur when standard software packages try to maximize biprobit function, even if number of covariates small. draw conclusions statistical practice. Finally, we prove conditions...
Summary Experimenters often use post-stratification to adjust estimates. Post-stratification is akin blocking, except that the number of treated units in each stratum a random variable because stratification occurs after treatment assignment. We analyse both and blocking under Neyman–Rubin model compare efficiency these designs. derive variances for post-stratified estimator simple difference-in-means different randomization schemes. nearly as efficient blocking: difference their order 1/n2,...
We provide a principled way for investigators to analyze randomized experiments when the number of covariates is large. Investigators often use linear multivariate regression instead simply reporting difference means between treatment and control groups. Their aim reduce variance estimated effect by adjusting covariates. If there are large relative observations, may perform poorly because overfitting. In such cases, least absolute shrinkage selection operator (Lasso) be helpful. study...
Summary Randomized controlled trials (RCTs) can provide unbiased estimates of sample average treatment effects. However, a common concern is that RCTs may fail to population We derive the assumptions are required identify effects from RCTs. placebo tests, which formally follow identifying and assess whether they hold. offer new research designs for estimating use non-randomized studies adjust RCT data. This approach considered in cost-effectiveness analysis clinical intervention: pulmonary...
The popularity of online surveys has increased the prominence using sampling weights to enhance claims representativeness. Yet, much uncertainty remains regarding how these should be employed in survey experiment analysis: they used? If so, which estimators are preferred? We offer practical advice, rooted Neyman–Rubin model, for researchers working with experimental data. examine simple, efficient estimators, and give formulas their biases variances. provide simulations that as well real...
<b>Objective</b> To compare the cost effectiveness of three most commonly chosen types prosthesis for total hip replacement. <b>Design</b> Lifetime model with parameters estimated from individual patient data obtained large national databases. <b>Setting</b> English National Health Service. <b>Participants</b> Adults aged 55 to 84 undergoing primary replacement osteoarthritis. <b>Interventions</b> Total using either cemented, cementless, or hybrid prostheses. <b>Main outcome measures</b>...
Estimating causal effects under exogeneity hinges on two key assumptions: unconfoundedness and overlap. Researchers often argue that is more plausible when covariates are included in the analysis. Less discussed fact covariate overlap difficult to satisfy this setting. In paper, we explore implications of observational studies with high-dimensional formalize curse-of-dimensionality argument, suggesting these assumptions stronger than investigators likely realize. Our innovation how strict...
A confidence sequence is a of intervals that uniformly valid over an unbounded time horizon. Our work develops sequences whose widths go to zero, with nonasymptotic coverage guarantees under nonparametric conditions. We draw connections between the Cram\'er-Chernoff method for exponential concentration, law iterated logarithm (LIL), and sequential probability ratio test -- our are time-uniform extensions first; provide tight, characterizations second; generalize third settings, including...
In cost-effectiveness analyses (CEA) that use randomized controlled trials (RCTs), covariates of prognostic importance may be imbalanced and warrant adjustment. CEA non-randomized studies (NRS), the selection on observables assumption must hold for regression matching methods to unbiased. Even in restricted circumstances when this is plausible, a key concern how adjust imbalances observed confounders. If propensity score misspecified, matched sample will imbalanced, which can lead...
There is an active debate in the literature on censored data about relative performance of model based maximum likelihood estimators, IPCW-estimators, and a variety double robust semiparametric efficient estimators. Kang Schafer (2007) demonstrate fragility IPCW-estimators simulation study with positivity violations. They focus simple missing problem covariates where one desires to estimate mean outcome that subject missingness. Responses by Robins, et al. (2007), Tsiatis Davidian Tan...
There is increasing interest in experiments where outcomes are measured by surveys and treatments delivered a separate mechanism the real world, such as mailers, door-to-door canvasses, phone calls, or online ads. However, common designs for often prohibitively expensive, vulnerable to bias, raise ethical concerns. We show how four methodological practices currently uncommon have previously undocumented complementarities that can dramatically relax these constraints when at least two used...