mthompson@utstat.toronto.edu
I am a recent graduate of the University of Toronto statistics Ph.D. program. The goal of my research was to make it possible for anyone performing statistical analysis to be able to use whichever model they believe is best for their data. To that end, I tried to develop Markov chain Monte Carlo methods that perform well on a wide variety of distributions with minimal tuning. I also developed ways to compare MCMC methods so that I and other researchers can better understand the performance and generalizability of methods we create.
I now work as a software engineer in the search industry, where I analyze the ways that people use search engines.
| 2011–08–24 | Introduction to SamplerCompare, a paper describing SamplerCompare. |
| 2011–08–24 | SamplerCompare 1.2.1, an R package for comparing the performance of MCMC samplers. |
| 2011–08–15 | Slice Sampling with Multivariate Steps, my Ph.D. thesis. (abstract, code for gridded radial steps) |
| 2010–11–16 | Graphical Comparison of MCMC Performance, a technical report describing the method for comparing the efficiency of MCMC methods used by SamplerCompare. |
| 2010–11–11 | Slice Sampling with Adaptive Multivariate Steps: The Shrinking-Rank Method, a manuscript for a talk I gave at the 2010 JSM. The slides for the talk and the code for the Gaussian process example are also available. |
| 2010–10–29 | A Comparison of Methods for Computing Autocorrelation Time, a technical report comparing four methods for computing autocorrelation time. One can replicate the results of this report with the R package ACTCompare 1.0. |
| 2010–03–08 | Covariance-Adaptive Slice Sampling, a technical report with my advisor, Radford Neal. (abstract, example software) |