Scientists across numerous disciplines attempt to identify and document causal relationships. Those unable to design and implement randomized (control) experiments to collect experimental data must resort to observational studies. To make causal inferences outside the experimental realm, researchers attempt to post-process large observational data sets to mimic experimental data. Finding the subset of data that closely resembles experimental data is a challenging problem. The primary approach used over the past thirty years has been to individually match treated and untreated (control) units to achieve a measure of balance.
This talk presents Mutual Information Matching (MIM), an optimization-based approach to performing group matching. To ensure that across the compared groups of units under study, treatment is independent of the units' covariates, matching models and algorithms are developed that focus on minimizing mutual information between the covariates and the treatment variable. Computational experiments demonstrate that MIM is very fast - running in close-to-linear time in the dataset size, - and achieves high-accuracy treatment effect estimates in comparison with the best previously existing methods.
Alexander Nikolaev is an Assistant Professor of Industrial and Systems Engineering, and Director of Social Optimization Laboratory at the University at Buffalo (SUNY). He received an M.S. degree in Applied Physics and Mathematics from Moscow Institute of Physics and Technology in 2001, an M.S. degree in Industrial Engineering from the Ohio State University in 2003, and a Ph.D. degree in Industrial Engineering from the University of Illinois at Urbana-Champaign in 2008. His research interests include machine learning, social network analysis, E-Learning, and E-Health. He has been a recipient of three National Science Foundation awards, and has published 34 papers in high-quality journals including JMLR, JMIR, Operations Research, OMEGA: International Journal of Management Science, and Transportation Research.
Date and time: Tuesday, August 22, 12:30 – 13.30
Location: De Grote Zaal (Kamer van Koophandel), J.F. Kennedylaan 2, 5612 AB Eindhoven
12.30-13.30 Lecture by Assistant Professor Alexander Nikolaev