Seminar Vol. 101

Title: Decision Making and Roc Curves

Speaker: Professor Han Hong, Stanford University

Time: June 7th, 2018 17:00–18:30

Venue: Conference Room 106B, Zhonghui Building (College of Economics, JNU)

Abstract:

The Receiver Operating Characteristic (ROC) curve is a representation of the information discovered by a statistical procedure in binary classification problem and is a key concept in machine learning and data science. This paper studies the statistical properties of ROC curves and its implication to model selection. We analyze the implications of different information asymmetry and incentive heterogeneity models on the relation between human decisions and the machine ROC curves. Our theoretical discussion is illustrated in the context of a large national level data set of pregnancy outcomes and doctor diagnosis from a survey data set of Pre-Pregnancy Checkups of reproductive age couples provided by the Chinese Ministry of Health.