The Simply Statistics Unconference on the Future of Statistics
Simply Statistics http://simplystatistics.org/ held an unconference on the future of statistics that was live streamed on Youtube October 30th, 2013 12pm-1pm EDT. The conference featured some of the brightest young minds discussing their views on where our discipline is going and had a 24 hour viewership of over 3,000. You can watch the whole unconference at http://bit.ly/198GoL3 or follow the real-time discussion on Twitter by searching the hashtag: #futureofstats. Below is a brief summary of each speaker’s points compiled by the Simply Statistics Bloggers Jeff Leek, Roger Peng, and Rafa Irizarry.
The Future of Statistical Software by Hadley Wickham
- The future of statistical software development will be on Github.
- We will move from static (PDF, LaTeX) to Web (R markdown, HTML) statistical documents.
- It is important to balance ease of use (R) and computational speed (Julia, C++) in statistical software.
The Future of Statistical Methods by Daniela Witten
- The future of statistical methods is renewed interest in inference.
- We are getting better at prediction; inferential understanding of machine learning is the future.
- We will move from black box machine learning to understanding scientific relationships.
The Future of Statistical Education by Joe Blitzstein
- The future of statistical education will be more applied and more computational and will involve clear statistical motivations for individual topics
- We will move away from overused data sets to using new and “live” data
- The future in some classes is moving the emphasis from basic calculus computations to deeper understanding.
The Future of Statistics in Biology by Hongkai Ji
- The future of statistics in biology will be determined by selection pressure of real problems
- There will be two equally important roles of statisticians: as safeguards and engines of discovery
- The future involves reanalysis of large public data sets to create new biological discoveries
The Future of Statistics in the Social Sciences by Sinan Aral
- The future of statistics in social sciences is designed experiments and causal inference
- The data we will observe will not be i.i.d. - neither are the data we had previously
- We will need sampling strategies that are scalable and deal with n.i.i.d. data; an example is the important issue of interference
The Future of Statistics and Data Science by Hilary Mason
- The future of statistics will be statistics done by everyone - people with statistics degrees and people without statistics degrees
- To thrive in industry, statisticians need to augment their stats knowledge with knowledge about software development in teams and how to engineer data products
- You often run into problems you’d like to solve in a year - but you only get a week - statisticians will need to be problem solvers.