David B. Dunson

Image of David B. Dunson

Arts and Sciences Professor of Statistical Science

Development of Bayesian statistical methods and approaches for uncertainty quantification motivated by applications with complex and high-dimensional data.  A particular interest is in high-dimensional low sample size data in which it is necessary to incorporate dimensional reduction through carefully designed prior distributions and challenges arise in efficiently computing posterior approximations. Ongoing focus areas include new algorithms for approximating posterior distributions in big data settings, nonparametric Bayes probability modeling allowing for uncertainty in distributional assumptions, analysis of network data, incorporating physical and geometric prior knowledge in modeling and novel models for dimension reduction for "object data" (functions, tensors, shapes, etc).  Primary application areas include genomics, neurosciences, epidemiology, and reproductive studies but with much broader interests in developing new methods motivated by difficult applications (in art, music, radar, imaging processing, etc).

Appointments and Affiliations
  • Arts and Sciences Professor of Statistical Science
  • Professor of Statistical Science
  • Professor in the Department of Electrical and Computer Engineering
  • Professor in the Department of Mathematics
  • Faculty Network Member of the Duke Institute for Brain Sciences
Contact Information:
Education:

  • Ph.D. Emory University, 1997
  • B.S. Pennsylvania State University, 1994

Research Interests:

Development of Bayesian methods motivated by applications with complex and high-dimensional data. A particular focus is on nonparametric Bayes approaches for conditional distributions and for flexible borrowing of information. I am also interested in methods for accommodating model uncertainty in hierarchical models, and in latent variable methods, including structural equation models. A recent interest has been in functional data analysis.

Specialties:

Bayesian Statistics
Complex Hierarchical and Latent Variable Modelling
Nonparametric Statistical Modelling
Model Selection
Statistical Modeling

Courses Taught:
  • STA 360: Bayesian Inference and Modern Statistical Methods
  • STA 393: Research Independent Study
  • STA 601: Bayesian and Modern Statistical Data Analysis
  • STA 790: Special Topics in Statistics
  • STA 993: Independent Study

Representative Publications: (More Publications)
    • Kessler, DC; Hoff, PD; Dunson, DB, Marginally specified priors for non-parametric Bayesian estimation., Journal of the Royal Statistical Society: Series B (Statistical Methodology), vol 77 no. 1 (2015), pp. 35-58 [10.1111/rssb.12059] [abs].
    • Durante, D; Dunson, DB, Bayesian dynamic financial networks with time-varying predictors, Statistics & Probability Letters, vol 93 (2014), pp. 19-26 [10.1016/j.spl.2014.06.015] [abs].
    • Kundu, S; Dunson, DB, Latent factor models for density estimation, Biometrika, vol 101 no. 3 (2014), pp. 641-654 [10.1093/biomet/asu019] [abs].
    • Rodriguez, A; Dunson, DB, Functional clustering in nested designs: Modeling variability in reproductive epidemiology studies, The annals of applied statistics, vol 8 no. 3 (2014), pp. 1416-1442 [10.1214/14-AOAS751] [abs].
    • Dunson, DB, Comment, Journal of the American Statistical Association, vol 109 no. 507 (2014), pp. 890-891 [10.1080/01621459.2014.955988] [abs].