A5013430806- Spatial and Panel Data Analysis
- Monetary Policy and Economic Impact
- Advanced Causal Inference Techniques
- Global Health Care Issues
- Healthcare Policy and Management
- Statistical Methods and Inference
- Fiscal Policy and Economic Growth
- Global trade and economics
- Energy, Environment, Economic Growth
- Advanced Statistical Methods and Models
- Genetic Associations and Epidemiology
- Health Systems, Economic Evaluations, Quality of Life
- Statistical Methods and Bayesian Inference
- Birth, Development, and Health
- Smoking Behavior and Cessation
- Obesity, Physical Activity, Diet
- Health disparities and outcomes
- Intergenerational and Educational Inequality Studies
- Global Public Health Policies and Epidemiology
- Environmental Impact and Sustainability
- Economic and Environmental Valuation
- Climate Change Policy and Economics
- Efficiency Analysis Using DEA
- Statistical Methods in Clinical Trials
- Employment and Welfare Studies
University of Oxford
2001-2024
University of Bristol
2013-2023
Medical Research Council
2014-2023
University of Guelph
2023
Cemmap
2005-2021
MRC Epidemiology Unit
2018-2021
University College London
1997-2021
Institute for Fiscal Studies
2004-2019
Leverhulme Trust
2002-2019
MRC Integrative Epidemiology Unit
2017-2018
Mendelian randomization (MR) is a powerful tool in epidemiology that can be used to estimate the causal effect of an exposure on outcome presence unobserved confounding, by utilizing genetic variants are instrumental variables (IVs) for exposure. This has been extended multivariable MR (MVMR) two or more exposures outcome.We use simulations and theory clarify interpretation estimated effects MVMR analysis under range underlying scenarios, where secondary acts variously as confounder,...
We consider testing for weak instruments in a model with multiple endogenous variables. Unlike Stock and Yogo (2005), who considered problem where the rank of matrix reduced form parameters is near zero, here we reduction one parameters. For example, two-variable model, instrument asymptotics [Formula: see text] are two reduced-form equations, vector constants sample size. investigate use conditional first-stage text]-statistic along lines proposal by Angrist Pischke (2009) show that, unless...
Monte Carlo studies have shown that estimated asymptotic standard errors of the efficient two-step generalised method moments (GMM) estimator can be severely downward biased in small samples. The weight matrix used calculation GMM is based on initial consistent parameter estimates. In this paper it extra variation due to presence these parameters accounts for much difference between finite sample and variance utilises moment conditions are linear parameters. This estimated, resulting a...
The system GMM estimator for dynamic panel data models combines moment conditions the model in first differences with levels. It has been shown to improve on differenced terms of bias and root mean squared error. However, we show this paper that covariance stationary AR(1) expected values concentration parameters levels equations cross‐section at time t are same when variances individual heterogeneity idiosyncratic errors same. This indicates a weak instrument problem also equation We 2SLS...
Abstract Estimates from Mendelian randomization studies of unrelated individuals can be biased due to uncontrolled confounding familial effects. Here we describe methods for within-family analyses and use simulation show that family-based reduce such biases. We illustrate empirically how effects affect estimates using data 61,008 siblings the Nord-Trøndelag Health Study UK Biobank replicated our findings 222,368 23andMe. Both within family reproduced established lower BMI reducing risk...
There is a widespread belief that peer effects are important in charitable giving but little evidence on how donors respond to their peers. Analysing unique data set of donations online fund‐raising pages, we find positive and sizeable effects: £10 increase the mean past increases by £2.50, average. Donations both very large small amounts changes mode. We signal charity quality – our preferred explanation use information earlier decide what appropriate for them give.
R-squared measures of goodness fit for count data are rarely, if ever, reported in empirical studies or by statistical packages. We propose several based on various definitions residuals the basic Poisson regression model and more general models such as negative binomial that accommodate overdispersed data. The preferred measure is deviance residual. An application to health-care-service utilization measured counts illustrates performance usefulness measures.
The generalized method of moments (GMM) estimation technique is discussed for count data models with endogenous regressors. Count can be specified additive or multiplicative errors. It shown that, in general, a set instruments not orthogonal to both error types. Simultaneous equations dependent variable often do have reduced form which simple function the instruments. However, simultaneous model and binary only logically consistent when system triangular. GMM estimator used explaining number...
Using many moment conditions can improve efficiency but makes the usual generalized method of moments (GMM) inferences inaccurate. Two-step GMM is biased. Generalized empirical likelihood (GEL) has smaller bias, standard errors are too small in instrumental variable settings. In this paper we give a new variance estimator for GEL that addresses problem. It consistent under asymptotics and, weak asymptotics, larger than and consistent. We also show Kleibergen (2005) Lagrange multiplier...
For regression models other than the linear model, R-squared type goodness-to-fit summary statistics have been constructed for particular using a variety of methods. The authors propose an measure goodness fit class exponential family models, which includes logit, probit, Poisson, geometric, gamma, and exponential. This is defined as proportionate reduction in uncertainty, measured by Kullback-Leibler divergence, due to inclusion regressors. Under further conditions concerning conditional...
Instrumental variables (IVs) can be used to construct estimators of exposure effects on the outcomes studies affected by nonignorable selection exposure. Estimators that fail adjust for will biased and inconsistent. Such situations commonly arise in observational studies, but are also a problem randomized experiments noncompliance. In this article, we review IV which outcome is binary, consider links between different approaches developed statistics econometrics literatures. The implicit...
Instrumental variable estimates of causal effects can be biased when using many instruments that are only weakly associated with the exposure. We describe several techniques to reduce this bias and estimate corrected standard errors. present our findings a simulation study an empirical application. For latter, we effect height on lung function, genetic variants as for height. Our demonstrates that, weak individual variants, two-stage least squares (2SLS) is biased, whereas limited...