University of Washington, 1989, PhD (Biostatistics)
University of Waterloo, 1984, M.Math (Statistics)
Simon Fraser University, 1983, BSc (Mathematics)
My research interests include the design and analysis of clinical trials, methods for exploratory analysis of survival data and adaptive non-parametric regression.
Most of my collaborative research focuses on the design, analysis and conduct of therapeutic clinical trials. I lead the statistical activities related to Phase II and Phase III clinical trials in lymphoma for the Southwest Oncology Group (SWOG). Recently I have been investigating design and analysis methods for targeting patient subgroups appropriate for both Phase II and Phase III clinical trials.
I am interested in the study of adaptive regression methods and their application to data arising from clinical trials. I have developed extensions or alternatives to tree-based methods to yield simple prognostic decision rules in collaboration with Dr. John Crowley (Cancer Research And Biostatistics, CRAB). One example of such methodology is a strategy for constructing prognostic groups based on recursive refinement or peeling algorithms which allow for calibration of the risk group in terms of outcome or fraction of patients in the group.
In addition, I have recently developed an algorithm called Extreme Regression for constructing either high- or low-risk outcome groups in collaboration with Dr. Charles Kooperberg and James Moon (FHCRC). The model leads to a simple inverse regression function which is represented as a Boolean decision rule similar to tree models. However, unlike trees, the new method can control the fraction of subjects identified by the rule.
Some of my other work has included methods for analyzing gene-expression and SNP data. A successful project was Logic Regression, a method which constructs predictors based on Boolean combinations of binary feature variables (led by Charles Kooperberg and originally part of Ingo Ruczinski's dissertation).
I am also currently working on methods that allow specification of genetic structure into the high dimensional regression problem.