Events
Technology Seminar Series: Ke-Wei Huang
Speaker: Ke-Wei Huang
Location: Henry Kaufman Management Education Center, Room 4-80 Chase Manhattan Classroom
Date: Friday, June 5, 2020
ABSTRACT:
In this seminar, I will present two ongoing research projects about applying machine learning methods to improve statistical analysis for social science applications. The research objective of the first study is to use deep-learning method to construct test statistics. The application highlighted in this seminar is about applying computer vision on Q-Q plot to construct a new test statistic for normality test. The proposed method integrates four components based on deep learning: an image representation learning component of a Q-Q plot, a dimension reduction component, a metrics learning component that best quantifies the differences between two Q-Q plots’ image representation, and a new normality hypothesis testing process. Our experimentation results show that the new approach can outperform several widely-used traditional normality tests. The research objective of the second study is to apply machine learning methods to better control latent homophily in a network regression targeting at identifying peer influence effects. This second project proposed three methods to better control latent homophily. The first method uses network embedding to construct more latent variables that could approximate the information of latent homophily variables. This set of embedding variables are added into regression as control variables. The second method is inspired by Robinson’s (1988) double residual estimator for partial linear regression that can control nonlinear effects from latent homophily variables. The third method is to estimate peer influence directly by Gradient-Boosted Machine method. Our experimentation results show that embeddings can indeed provide information about latent homophily and all three methods could better control latent homophily when compared with existing regression methods. Preliminary theoretical analysis of two new estimation methods will be discussed.
BIO:
Dr. Ke-Wei Huang is an Associate Professor with the Department of Information Systems and Analytics at the School of Computing of National University of Singapore. Prior to joining NUS in 2007, he obtained his Bachelor’s degree in electrical engineering (1995) and M.B.A. in finance (1997) from National Taiwan University, M.S. degree (2002) and Ph.D. degree in Information Systems (2007) from Stern School of Business at New York University. His fields of research include using machine learning to improve social science research methods, machine learning applications in finance or accounting, entrepreneurship and innovation research for IT and knowledge workers, and pricing information goods. His works have been published in Information Systems Research, Strategic Management Journal, Production and Operations Management, Journal of Economics & Management Strategy, Quantitative Marketing and Economics, IEEE Transaction on Engineering Management, Decision Support Systems, and ACM Transactions on MIS.