Michael R. Elliott, PhD

Michael Elliott is a Professor of Biostatistics at the University of Michigan School of Public Health and Research Professor at the Institute for Social Research. He received his PhD in biostatistics in 1999 from the University of Michigan. Prior to joining the University of Michigan in 2005, he held an appointment as an Assistant Professor at the Department of Biostatistics and Epidemiology at the University of Pennsylvania School of Medicine. Dr. Elliott's statistical research interests focus around the broad topic of "missing data," including the design and analysis of sample surveys, causal and counterfactual inference, and latent variable models. He has worked closely with collaborators in injury research, pediatrics, women's health, the social determinants of physical and mental health, and smoking cessation research. Dr. Elliott has served as an Associate Editor for the Journal of the Royal Statistical Society, Series C and the Journal of the American Statistical Association, and as an Associate Editor and Editor of the Journal of Survey Statistics and Methodology. He was Associate Chair of Academic Affairs for the Department from 2018-2021.

BIOSTAT651: Theory And Application Of Generalized Linear Models

BIOSTAT855: Regression Models in Complex Sample Design Settings (JPSM/MPSM 895)  

BIOSTAT870: Analysis of Repeated Measurements

BIOSTAT617: Methods and Thoery of Sample Design

BIOSTAT681: Introduction To Causal Inference

• PhD, Biostatistics, University of Michigan, 1999

• M.S., Biostatistics, University of Michigan, 1997

• B.A., Mathematics, University of Chicago, 1985

• My methodological research focuses in two major areas: design and analysis of population-based surveys, and development of causal modeling estimators. Together, these areas may be coherently thought of as special cases of missing data: population surveys are censuses from which typically most of the population data is missing, while causal estimators, in particular "potential outcomes" models, can be viewed as analyses in which outcomes under different treatment assignments cannot, by design, be observed.

  My focus in the design and analysis of population-based surveys has been on the development of model-based Bayesian approaches that complement traditional design-based analyses of complex sample survey data. Traditional design-based approaches to analyzing survey data are non-parametric, but rely on asymptotic assumptions and can be highly inefficient. Model-based approaches can have better small sample and efficiency properties, but often lack the robustness of design-based methods. Modern techniques such as non-parametric regression and Dirichlet processes allow development of models than can balance robustness-efficiency tradeoffs. This work is also important as access to administrative and other non-probability samples increases the need for methods to combine probability and non-probability samples.
  
  In the causal modeling arena, I have considered links between the Rubin Causal Model (RCM), which posits "principal strata" formed by pre-randomization counterfactual compliance behavior, and the marginal structural mean model proposed by Robins and others, as well as explored extensions of the RCM to longitudinal randomized trials where patients have failed to adhere to their randomization arm.  This work can be generalized into the surrogate marker setting, where it is important to determine to what degree easy-to-observe biomarkers are on causal pathway for outcomes that are expensive or time-consuming to observe.  Similarly I am pursuing work looking at developing methods to adjust for bias due to “treatment by indication” in observational settings, where preliminary outcomes may drive treatment decisions. I have also begun to look at methods to improve the generalizability of clinical trials by incorproating information from relevant probability samples.

  Beyond these two areas, I have also considered other latent variable methodology problems, developing extensions of generalized growth curve mixture models in the context of psychiatric affect data. Most recently I have started to consider whether models that focus on variability structures rather than, or in addition to, mean structures, might be useful in this or other analytic settings (e.g., hormone and menstruation time series data).

  I have a wide variety of applied interests. I have worked a great deal on pediatric issues, including serving as the lead biostatistician on the Partners for Child Passenger Safety, which has conducted research into the cause and prevention of injuries to children in passenger vehicles, and now as the lead biostatistician on the Michigan cohort for the Environmental Influences on Child Health Outcomes (ECHO), a national study of children for the National Institutes of Health. I have worked in women's health issues, including studies designed to understand the onset of menopause and to predict and ultimately treat health problems that accompany the menopausal transition.  I also collaborate with researchers at the Institute for Social Research on environmental effects on health outcomes, including factors such as work stress and built environment on health in later life. Finally, I collaborate with researchers at the University of Michigan Transportation Research Institute on a wide variety of topics related to driving behavior and driver safety, including work on a large naturalistic driving study funded though the Michigan Institute for Data Science  (MIDAS). I am also collaborating with smoking cessation researchers, in particular to assess the impact of various legal restrictions on smoking behavior.

• Tan, Y. V., Flannagan, C.A.C., Elliott, M.R. (2021). "Accounting for selection bias due to death in estimating the effect of wealth shock on cognition for the Health and Retirement Study," Statistics in Medicine, 40, 2613-2625.

• Elliott, M.R., Zhao, Z., Muhkerjee, B., Kayna, A., Needham, B.L. (2020).  “Methods to Account for Uncertainty in Latent Class Assignments when using Latent Classes as Predictors in Regression Models, with Application to Acculturation Strategy,”  Epidemiology, 31, 194-204.

• Zhou, T., Elliott, M.R., Little, R.J.A. (2019). “Penalized Spline of Propensity Methods for Treatment Comparison,” Journal of the American Statistical Association (with discussion), 114, 1-38.

• Elliott, M.R., Valliant, R. (2017).  Inference for Non-probability Samples. Statistical Science, 32, 249–264.

• Zhou, H., Elliott, M.R., Raghunathan, T.E. (2016). A Two-Step Semiparametric Method to Accommodate Sampling Weights in Multiple Imputation. Biometrics 72, 242-252.

• Jiang, B., Elliott, M.R., Sammel, M.D., Wang, N. (2015). Joint Modeling of Cross-Sectional Health Outcomes and Longitudinal Predictors via Mixtures of Means and Variances. Biometrics, 71, 487-497.

• Elliott, M.R., Conlon, A.S.C., Li, Y., Kaciroti, N., Taylor, J.M.G. (2015). Surrogacy Marker Paradox Measures in Meta-Analytic Settings. Biostatistics, 16, 400-12.

• Elliott, M.R., Sammel, M.D., Faul, J. (2012). Variability as a Predictor of Health in Longitudinal Studies. Statistics in Medicine, 31, 2745-2756.

• Elliott, M.R., Raghunathan, T.E., Li, Y. (2010). Bayesian Inference for Causal Mediation Effects Using Principal Stratification with Dichotomous Mediators and Outcomes. Biostatistics, 11, 353-372.

• Lin, J.Y., Ten Have T.R., Elliott, M.R. (2009). Nested Markov Compliance Class Model in the Presence of Time-Varying Noncompliance. Biometrics, 65, 505-513.

• Elliott, M.R. (2009). Model Averaging Methods for Weight Trimming in Generalized Linear Regression Models. Journal of Official Statistics, 25, 1-20.

• Elliott, M.R., Little, R.J.A. (2005). A Bayesian Approach to 2000 Census Evaluation using A.C.E. Survey Data and Demographic Analysis. Journal of the American Statistical Association, 100, 380-388.

• International Statistical Institute
• American Statistical Association
• Royal Statistical Society
• ENAR, International Biometric Society
• Association for the Advancement of Automotive Medicine