Interview with Hal Varian: Chief Economist at Google; Emeritus Professor at the University of California, Berkeley in three departments; business, economics, and information management.
C&S: What are the biggest challenges for governance/regulation created by growth of the big data market? Are there big differences between the US/Chinese and European approaches to big data opportunities?
Hal VARIAN: There are policy issues relating to data access and control that arise constantly. This generates a lively debate, to say the least. As an economist, I would like to see serious benefit-cost analysis guide regulatory policy.
What are the most important skills sets for those who need to make sense of results of big data analytics?
Statistics and machine learning are most obvious. But in order to put analysis to work, communication skills are critically important. To be effective, a data analyst needs to turn data into information, information into knowledge, and knowledge into action. You can't do this without communication.
What are the biggest opportunities for business and are businesses able to make effective use of big data to improve their margins?
As in every business, it is imperative to understand your customer. When you can draw on computer mediated transactional data, it is possible to gain a deeper understanding of the customers' needs than was previously the case.
What has big data analytics to learn from mainstream econometrics and what can big data analytics contribute to mainstream econometrics?
Econometrics can draw on some of the powerful techniques of predictive analytics that have been developed by the machine learning community. These tools are particularly helpful when dealing with data involving nonlinearities, interactions, and thresholds.
Econometrics, on the other hand, has focused on causal inference from its very early days. Techniques such as instrumental variables, regression discontinuity, and difference-in-differences have been widely used in econometrics but, to date, have not been used in the machine learning community.
Finally, the statistical field of experimental design will be valuable to both communities, as computer mediated transactions enable true randomized treatment-control experiments, which are the gold standard for causal inference.
What should be added to standard US Ph.D. programs in economics to make the students big data literate?
There are now very good textbooks, online tutorials, and tools that make it relatively easy to put together a course on machine learning. In addition virtually all computer science departments and many statistics departments offer such courses.
|Printer friendly Cite/link Email Feedback|
|Publication:||Communications & Strategies|
|Date:||Jan 1, 2015|
|Previous Article:||Interview with Damien Cudel: data insight product Marketing Manager, Microsoft.|
|Next Article:||Desperately looking for a data ethic: the importance of a Universal declaration on digital human rights.|