Collaborative Research: Computational Environment for Bayesian Inference in the Social Sciences

Harvard University, Cambridge MA

Investigators

Abstract

The advent of Markov chain Monte Carlo (MCMC) methods is the most important development in statistical computing within the last fifteen years. While these algorithms have allowed statisticians to fit almost any conceivable model, statisticians have been (with notable exceptions) the only people who have been able to take full advantage of these estimation methods. This project provides a computational environment that puts MCMC methods in the hands of social scientists so that they too can use the power of these algorithms to fit innovative statistical models of their choosing. The project provides free, open-source, easy-to-use software for Bayesian inference that is geared towards the needs of social scientists. It also provides a documented development environment others can use to easily implement non-standard statistical models, and a mechanism for other researchers to distribute their own software with a consistent user interface and documentation. The project is based on a scientific approach to the provision and development of statistical software. The software development is cumulative and builds on the work of others; it is free, open-source, and cross-platform, thus allowing for widespread dissemination and an extremely quick development-release cycle. Further, because all users access to the underlying source code, it is straightforward to fix bugs and extend the software. The project puts powerful statistical methods into the hands of empirically-oriented social and behavioral scientists. As a result, it will improve the quality of empirical work done in these areas of study. More specifically, the project develops and provides computer software for statistical learning. The software makes use of state-of-the-art algorithms while remaining easy to use. The project also contains an instructional component that provides a suite of demonstration programs for use in undergraduate and graduate teaching. Furthermore, students will be directly benefited through first-hand involvement in the project.