Ridge Estimator for Linear Instrumental Variables using K-fold Cross-Validation
Speaker: Fallaw Sowell, Carnegie Mellon University, USA
DATE: Wednesday 15 November 2023
TIME: 13.00 – 14.00
6, Rue Richard Coudenhove-Kalergi
– Free seminar
– Registration link: as of October 30th, 2023
Tel: +352 46 66 44 6283
This talk discusses issues caused by regularization in a causal inference task (instrumental variables) and potential fixes. First, the distinction between business problems answered with predictions versus business problems answered with structural models and causal inference is noted. Prediction problems frequently use machine learning tools that require selecting a regularization parameter. The optimal model can be used for predictions, but its sampling distribution is rarely considered. Alternatively, structural models in causal inference focus on the parameters’ signs and statistical significance. The estimated parameters can guide how to control the system to achieve a desired outcome. The talk then presents recent results in estimating structural models for causal inference that use regularization. The challenge is accounting for the empirically selected tuning parameter in the sampling distribution of the parameters of the structural model. An additional challenge for the sampling distribution is that the tuning parameter must be restricted to nonnegative values and converge to zero, which is on the boundary of the parameter space.
About Fallaw Sowell:
Fallaw Sowell is an Associate Professor of Economics at the Tepper School of Business. His research interests include causal inference, machine learning, time series, and applied microeconomics. He was an associate editor for the Journal of Econometrics for 20 years. He has consulted with the Postal Rate Commission, the Refractories Institute, Trane/American Standard Inc., and West Chester Capital. He has worked for the Federal Reserve Board of Governors and the Commodities Futures Trading Commission. He has served as the Deputy Dean of Master’s Programs for the Tepper School of Business. He has been the Director of the MBA program and the Director of the Master’s degree in Computation Finance at Carnegie Mellon University. His Ph.D. in Economics is from Duke, and his M.S. in Statistics is from the University of North Carolina, Chapel Hill. He likes college basketball and fly fishing.
Supported by the Luxembourg National Research Fund (FNR) 17931929