Online Causal Inference Seminar

Free and open to the public

Tuesday, April 5, 2022 [Link to join] (ID: 996 2837 2037, Password: 386638)

• Speaker: Zijian Guo (Rutgers University)
• Title: Two Stage Curvature Identification with Machine Learning: Robust Causal Inference with Invalid Instrumental Variables
• Discussant: Frank Windmeijer (University of Oxford)

• Abstract: Instrumental variables regression is a popular causal inference method for endogenous treatment. A significant concern in practical applications is the validity and strength of instrumental variables. This paper plans to make inferences for the treatment effect when all instruments are possibly invalid. To do this, we propose a novel two stage method and a generalized concept to measure the strengths of possibly invalid instruments: such invalid instruments can still be used for inference in our framework called two stage curvature identification (TSCI). In the first stage, we fit the treatment model with a machine learning method and encode the model by a transformation matrix; in the second stage, we leverage the transformation matrix to estimate the treatment effect by adjusting violation forms of instrumental variables. We propose a novel bias correction method to remove the overfitting bias from machine learning methods. Among a collection of spaces of violation functions, we choose the best one by evaluating invalid instrumental variables' strength and comparing the TSCI estimators given by various spaces. We demonstrate our proposed TSCI methodology over a large-scale simulation and apply our method to estimate the effect of education on earnings and the class size on scholastic achievement. This is a joint work with Dr. Peter Bühlmann.