Learning to Love Bayesian Statistics – Allen Downey ODSC Boston 2015
ConferencesModelingStatisticsODSC East 2015|Speaker Slidesposted by Open Data Science June 3, 2015 Open Data Science
Bayesian statistical methods provide powerful tools for answering questions and making decisions. For example, the result of Bayesian analysis is a set of values and probabilties that can be fed directly into a cost-benefit analysis, which is not possible with conventional statistics.
But there are several barriers to widespread adoption: people with the knowledge and ability to apply these methods to practical problems are rare, and there are few accessible resources for developing these skills.
In this presentation, I will explain the advantages of Bayesian methods over classical approaches using concrete examples like A/B testing. I will recommend resources and suggest steps data science teams can take to develop skills and begin applying Bayesian methods to real-world problems.
What’s wrong with classical statistics?
How do Bayesian methods address these problems?
What are the challenges?
How do we get started?
How do these methods scale up in production?
Professor Allen Downey teaches classes in software engineering and data science as well as physical modeling and simulation. He is the author of several textbooks in use at Olin and many other schools; they include Think Python, Think Stats, and Physical Modeling in MATLAB. These books are available under free licenses that allow readers to copy and modify the text as well as contribute material. In 2012 Prof Downey received the Undergraduate Computational Engineering and Sciences (UCES) Award for developing an innovative undergraduate class on Complexity Science and an accompanying book, Think Complexity. In 2009-10 he was a Visiting Scientist at Google, Inc., working in their network infrastructure group on projects related to the Make the Web Faster initiative. Before coming to Olin, Prof Downey taught at Colby College and Wellesley College, and held research positions at the San Diego Supercomputer Center and Boston University. He received his Ph.D. in computer science from the University of California/Berkeley in 1997, with a dissertation on operating system support for large-scale parallel computation. His undergraduate and master’s degrees are from the Civil Engineering department at MIT.