Recent work with François Caron and Arnaud Doucet has been posted on ArXiv. In the paper we propose an original class of flexible Bayesian linear models for dynamic sparsity modelling. The proposed class of models expands upon the existing Bayesian literature on sparse regression using generalized multivariate hyperbolic distributions. The properties of the models are explored through both analytic results and simulation studies. We demonstrate the model on a financial application where it is shown that it accurately represents the patterns seen in the analysis of stock and derivative data, and is able to detect major events by filtering an artificial portfolio of assets.

Recent work with Gavin Shaddick and Jim Zidek, titled "Modelling Nonstationary Processes Through Dimension Expansion," is set to appear in the Journal of the American Statistical Association. The proposed method works by expanding the geographic plane over which these processes evolve into higher dimensional spaces, transforming and clarifying complex patterns in the physical plane. By combining aspects of multi-dimensional scaling, group lasso, and latent variables models, a dimensionally sparse projection is found in which the originally nonstationary field exhibits stationarity. Can't wait for the article to appear? Check out a preprint on ArXiv.

Several years after a project initiated by Agriculture and Agri-Foods Canada kicked off, Jim Zidek and I have published a paper detailing our methodology for efficiently predicting crop yields in the Canadian Prairies. The paper has been published in Agriculture and Forest Meteorology, and you can find it on the publications page.

Ongoing work with Pierre Jabob, Pierre Del Moral, and Arnuad Doucet is now available as a pre-print on arXiv. The paper builds upon the framework of the Wang-Landau algorithm to build an automated density exploration algorithm for general purpose application. An R package to accompany the paper is also available for download. This code implements the Wang-Landau algorithm along with the numerous improvements proposed in the paper.

Some work completed jointly with Francois Caron is now published in Bayesian Analysis. The paper, titled "Bayesian Clustering in Decomposable Graphs," is available for download from the journal's website. The paper develops a novel class of prior distributions over decomposable graphs which allow for vastly improved modeling flexibility over existing options.

After nearly 4 years in the revision process, some of the work coming out of my 6-month visit to Los Alamos National Labs in 2007 is now being published in Engineering Structures, a top-tier structural engineering given an A* rating in the 2010 ERA journal ranking. The paper, with Sezer Atamturktur and Francois Hemez, focuses on using time and frequency domain techniques to detect damage and anomalies in masonry monuments, which due to their unique size, status, and age require a unique statistical approach to diagnose anomalies in the presence of massive environmental variation. You can find the paper on the publications page.

Recent work with Arnaud Doucet and Raphael Gottardo examining sequential Monte Carlo as a tool for efficiently conducting prior sensitivity and cross-validation will appear in the coming Canadian Journal of Statistics. You can find a preprint here.
 
 
 
 

In the nearly 3 years which have passed since presenting a poster at the CMS/MITACS conference entitled "The Bayesian Elastic Net", the interest in the topic has grown so considerably that we have revived our work in a more easily available format. Hence we have created a tech report entitled "Grouping Priors and the Bayesian Elastic Net" which details our original work. You can find the manuscript here or on arXiv.