Steve MacEachern

STUDENTS

Through time, I've had a chance to work with a wonderful batch of students. For a full list of those I've advised, click on the link above (Students). Recent and imminent graduates who may be interested in a position include those below:

• Yoonsuh Jung (2010) (with Yoonkyung Lee)
  Case-specific regularization methods
  Postdoctoral Researcher, M.D. Anderson Cancer Center
  

  Yoonsuh is currently working with Jianhua Hu and Xuming He as a postdoctoral fellow. His dissertation work blends the ideas of regularization (e.g., penalized likelihood) with the treatment of individual cases in a modelling framework to enhance robustness or efficiency of estimators. His dissertation work will soon appear in Statistical Science and also appears on his web page.

• Ju Hee Lee (2010)
  Robust statistical modeling through nonparametric Bayesian methods
  Postdoctoral Researcher, M.D. Anderson Cancer Center
  

  Juhee is currently working with Peter Mueller and Yuan Ji as a postdoctoral fellow. Her research on the postdoc involves hierarchical Bayesian modelling of gene expression data and also local clustering methods. The work is connected to nonparametric Bayesian techniques, and it has strong applied and computational elements.

• Pingbo Lu (projected graduation, spring 2012) (with Xinyi Xu)
  Calibrated Bayes factors for model selection and model averaging

  Pingbo is examining the unusual behavior of the Bayes factor--first favoring one model to an extreme degree, then flipping to favor the other model to an extreme degree--that is so often observed in practice when comparing models of very different size. The problem is particularly severe when the models are high or infinite dimensional (e.g., nonparametric Bayes) and the prior distribution is specified by a fairly arbitrary rule.

• Hang Joon Kim (projected graduation, spring 2012)
  Bayesian models and computation for consumer choice

  Hang has completed work on a generalization of the "multiset sampler" (a strikingly innovative MCMC algorithm due to Scot Leman and coauthors), improving performance of the basic algorithm and placing it squarely into the framework of existing MCMC methods. He is currently using the algorithm to fit random effects models, models for outliers, and a number of other tricky-to-fit models.