It is virtually impossible to separate modern statistics from computing. The statistical models we develop keep getting more sophisticated and the kinds of data we analyze are larger and more complex than ever before. This has led to computational statistics – the study of algorithms designed to solve computational problems that arise in statistics – being a discipline of its own within statistics.
At Penn State Statistics, we have a history of seminal contributions to computational statistics, from the composite likelihood methods developed by Bruce Lindsay in the 1980s to the MM (minorization-maximization) algorithms developed by David Hunter in the late 1990s. We have several faculty who work on developing new computational methods for solving challenging statistical problems. Often they work on other research that motivates these problems, such as network models, spatial statistics, astronomy, functional data, and high-dimensional regression.
The computational research expertise in our department includes the following areas: MM and EM (expectation-maximization) algorithms (David Hunter and Jia Li), Markov chain Monte Carlo and other algorithms for approximating high-dimensional integrals and performing Bayesian inference (Stephen Berg, Hyungsuk Tak, Murali Haran, Ephraim Hanks, Le Bao), optimization in the context of high-dimensional data (Lingzhou Xue, Runze Li), algorithms and methods in the context of machine learning (Jia Li, Bharath Sriperumbudur). Several of our faculty also have strong ties to the Institute of Computational and Data Sciences (ICDS) which is a cross-college institute that supports computing research through infrastructure and opportunities for developing research collaborations.