Dr. Justin Jia
Department of Business Analytics & Statistics
The University of Tennessee
Friday, October 25, 2019 2:30-3:30pm
In a prescriptive analytics problem, a decision maker aims to solve an optimization problem involving certain random variables. The decision maker does not know the distribution of the random variables, but instead observes a sample from the distribution and makes a decision based on the sample. In this study, we develop a solution framework and show that for a practical class of problems, the decision, a high-dimensional function of data, can be reduced to a simple single-dimensional function of a “sufficient prescriptive statistic.” We fully characterize the solution and derives its asymptotic properties and finite-sample performance guarantees. Applying the solution procedure to data-driven auction design, we obtain insights on the impact of data on decisions. We further report on a laboratory experiment with financially incentivized human subjects and show that behavior qualitatively matches theoretical predictions.
Justin Jia is an assistant professor in the Department of Business Analytics and Statistics at the University of Tennessee. He holds a Ph.D. in Supply Chain Management from Penn State University and served as a faculty member at Purdue University before joining UT. His research interests lie in the areas of prescriptive analytics, healthcare supply chain, e-operations and inventory management. He has published articles in journals such as Production and Operations Management and Management Science.