My research examines the fundamental and practical differences between using statistical and other empirical methods for prediction compared to causal explanation and to description. Although the discussion of explanation vs. prediction has been actively pursued in the philosophy of science, the statistics literature has not considered it in a holistic way. Yet, statistical modeling can be and is used for each of these goals.
To Explain Or To Predict travels to Prague! I'll be delivering a keynote address at the upcoming Discovery Summit on Wednesday March 22, 2017, 11AM at the Prague Marriott Hotel. The two other keynote speakers are John Sall (co-founder and executive VP of SAS) and Prof. Ron Kenett.
Next week I'll be headed to SAS headquarters in Cary, NC for an interview on Analytically Speaking. I look forward to Anne Milley's thought-provoking questions! The announcement promises to walk through several of my research areas:
On Wednesday (March 23) I'll talk about the difference between using statistical models and data mining for explanation vs. prediction at NTHU's Institute of Information Systems & Applications (here's the official announcement). For those who've been waiting to hear this talk in Hsinchu, here's your chance. Looking forward to a lively discussion.
Location: NTHU, Delta Building, Room 105
Date & Time: March 23, 13:30-15:00
On Friday, March 4 (2016), I'll deliver a talk on Big Data - To Explain or To Predict as part of the Big Data Experts Speaker Series @ Rotman School of Management, University of Toronto. The talk will discuss the differences between modeling data for causal explanation vs. prediction, with the aim of clarifying usages of big data analytics in both academic research and industry.
The Microsoft Research group in NYC invited me to give a talk on "To Explain or To Predict? How Prediction Can Advance Research". I spent half a day on Nov 3, 2015 at their beautiful lab and learned what "computational social scientists" study. The audience in my talk included folks with a computer science background from the computational social science and machine learning groups and others.
Partial Least Squares Path Modeling (PLS-PM) is a popular statistical modeling tool in information systems, marketing and other social science disciplines. Researchers using PLS-PM have typically focused on explanatory modeling. I will be discussing PLS from a predictive point of view at two upcoming conferences this month: