Wednesday, April 9, 2014

Trusting Data Science

Victoria is awesome. I guarantee she will impress you. If you are interested in the future of social science research, you should go to this seminar! PLUS, there's a free lunch if you RSVP!

UC Berkeley and Stanford are pleased to present:

Data, Society, and Inference Seminar
When Should We Trust the Results of Data Science?
Department of Statistics, Columbia University

Monday, April 14, 1pm – 2:30pm
Blum Hall 330, UC Berkeley (map)

Please join us for another session of the Data, Society, and Inference seminar. To help us coordinate lunch, please RSVP by clicking here.

When should we trust the results of data science? In this talk, I will take a critical view of knowledge generation in data science. I develop three lines of thought that point to the need for new methods for reliable inference in computational science. First, the use of computational methods facilitates enormously complex calculations that are not well-described in a traditional scientific publication. Verification and validation of these findings will typically require access to the computer codes used, as well as the data upon which these calculations are based. Second, traditional and commonly used methods of statistical inference can be misleading, or even inappropriate in big data settings. Finally, new thinking around research processes that increase the reliability of computational findings will be presented. I will conclude by discussing efforts to address each of these issues.

No comments:

Post a Comment