David Todem, MSPH, PhD

Professor of Epidemiology and Biostatistics

PhD in Statistics (2003),
University of Wisconsin, Madison
MS in Population Health (2002),
University of Wisconsin, Madison
MS in Biostatistics (1998),
Hasselt University, Belgium   

Department of Epidemiology and Biostatistics
909 Wilson Road  Room B601
Michigan State University
East Lansing 48824
517 353.8623

David Todem, MSPH, PhD is a Professor of Biostatistics in the Department of Epidemiology & Biostatistics, and an Adjunct Associate Professor in Statistics & Probability at Michigan State University (MSU). His current line of research focuses on the development of statistical methods to analyze data generated from population-based studies pertaining to Alzheimer’s disease, Cancer and Dental Caries. Specifically, his primary research interests in methods evolve around the following topics: 1) Models for longitudinal, clustered and functional data; 2) Joint models for (multi-state) time-to-event endpoints and longitudinal outcomes; 3) Inferential techniques for non- and weakly-identifiable models with application to informative nonresponse; and 4) Testing procedures for evaluating non-negative heterogeneity parameters with application to mixture models.

One interesting application of longitudinal and multi-state models is the study of temporal patterns of promising biomarkers and brain imaging parameters to understand the pre-clinical stages of Alzheimer's disease and its progression over time. This methodological work is motivated by the Alzheimer's disease neuroimaging initiative (ADNI1, ADNI-GO and ADNI2) study, a large public database well suited for longitudinal investigations on AD, due to its breath of serial data on biomarkers and cognitive markers. Another interesting application is in cancer research involving longitudinal data on quality of life endpoints for patients during and post treatment. Models for serial data on quality of life outcomes are critically important in understanding the effectiveness of a treatment regimen in cancer patients.

This methodological research has expanded to include methods for missing data, given the very nature of longitudinal studies to generate dropouts. Dr Todem is particularly interested in situations where the missing data process and attrition depend on the unobserved response. His analytical strategy has been to use a versatile approach that embeds the treatment of incomplete data in the context of sensitivity analysis that is informed by the subject matter. Within this framework, Dr Todem is particularly interested in developing methods for conducting inferences when the assumed working (non-ignorable) model is at best weakly identifiable in light of observed data.

Dr Todem is also interested in various aspects of mixtures models which provide a natural framework to describe heterogeneity in a population. A prevailing concern for this class of models is whether the inherent heterogeneity is consistent with observed data. In this line of research, Dr Todem is particularly interested in developing testing procedures for evaluating homogeneity against varying heterogeneity, a non-standard problem arising in dental caries research and other applications in epidemiology and medical research.

In addition to his methods research, Dr Todem collaborates actively with other researchers on various projects in epidemiology and medicine. Ongoing collaboration projects include: 1) a neuroimaging study on the neural development underlying childhood stuttering, 2) a study on Organochlorine and gene expressions of sex steroids in a multi-generational cohort, and 3) an intervention study aiming at evaluating a family-based cancer literacy model to increase participation in breast and cervical cancer control programs among medically underserved female population in the US.

Dr Todem’s methods works and collaborations are primarily supported by funded grants from the National Institutes of Health (NIH). He was a recipient of a Faculty Career Development K01 Award through the National Cancer Institute (2009-2015). This is the most prestigious of NIH's award designed to help top-performing researchers early in their careers to simultaneously develop their contribution and commitment to research and education.



(*current/former students’ paper with D. Todem serving as the corresponding author)

Todem, D., Hsu, W-W and Fine, J.P. (2017). A quasi-score statistic for homogeneity testing
against covariate-varying heterogeneity (invited revision)

Todem, D., Kim, KM and Hsu, W-W (2016). Marginal mean models for zero-inflated count data,
Biometrics 72, 986-994

*Hsu, W-W, Todem, D. and Kim, KM (2016). A sup-score test for the cure fraction in mixture
models for long-term survivors, Biometrics 72, 13481357

*Hsu, W-W, Todem, D. and Kim, KM (2014). Adjusted supremum score statistics for evaluating non-standard hypotheses, Scandinavian Journal of Statistics, 42:746-759

*Cao, G., Hsu, W-W. and Todem, D. (2014). A Score-type Test for Heterogeneity in Zero-inflated Models in a Stratified Population, Statistics in medicine, 33(12):2103-14.

*Cao, Q., Todem, D., Yang, LJ and Fine, J.P. (2013). Evaluating statistical hypotheses using weakly-identifiable estimating functions, Scandinavian Journal of Statistics, 40(2):256-273

Todem, D., Hsu, W.-W. and Kim, K. (2012). On the efficiency of score tests for homogeneity in two-component parametric models for discrete data. Biometrics, 68(3): 975-982

Cao, G., Yang, L. and D. Todem, (2012). Simultaneous inference for the mean function based on dense functional data, Journal of Nonparametric Statistics, 24(2):359-377

*Zhang, Y., Todem, D., Kim, K. and Lesaffre, E. (2011). Bayesian latent variable models for spatially correlated tooth-level binary data in caries research, Statistical Modelling, 11(1):25-47

Todem, D., Fine, J and Peng, L. (2010). A global sensitivity test for evaluating hypotheses with unidentifiable models, Biometrics, 66: 558–566

Todem, D, K. Kim, J.P. Fine and L. Peng (2010). Semi-parametric regression models and sensitivity analysis of longitudinal data with nonrandom dropouts, Statistica Neerlandica, 64(2):133-156

Todem, D, and Williams, K.P. (2009). A hierarchical model for binary data with dependence between the design and outcome success probabilities, Statistics in Medicine, 28:2967-2988

Williams, K.P, Mullan, P.B, Todem, D. (2009). Moving from Theory to Practice: Implementing the Kin Keeper Cancer Prevention Model, Health Education Research, 24(2):343-356

Romero, R., Nien, JK, Espinoza, J. Todem, D., Fu, W., Chung, H. Kusanovic, JP., Gotsch, F. Erez, O. Mazaki-Tovi, S. Gomez, R. Edwin, S., Chaiworapongsa, T., Levine, RJ, Karumanchi, S.A (2008). Longitudinal study of angiogenic (placental growth factor) and anti-angiogenic (soluble endoglin and soluble vascular endothelial growth factor receptor-1) factors in normal pregnancy and patients destined to develop preeclampsia and deliver a small for gestational age neonate. The Journal of Maternal-Fetal & Neonatal Medicine, 21(1):9- 23.

Romero, R., Nien, JK, Espinoza, J. Todem, D., Fu, W., Chung, H. Kusanovic, JP., Gotsch, F. Erez, O. Mazaki-Tovi, S. Gomez, R. Edwin, S., Chaiworapongsa, T., Levine, RJ, Karumanchi, S.A (2008). The change in concentrations of angiogenic and anti-angiogenic factors in maternal plasma between the first and second trimesters in risk assessment for the subsequent development of preeclampsia and small-for-gestational age, The Journal of Maternal-Fetal & Neonatal Medicine, 21(5): 279–287