A0541
Title: Testability of instrumental variables in additive nonlinear, non-constant effects models
Authors: Feng Xie - Beijing Technology and Business University (China) [presenting]
Abstract: The aim is to address the testability of instrumental variables derived from observational data. While much of the existing literature on testable implications deals with discrete treatments or assumes constant causal effects, real-world applications often involve continuous treatments (e.g., drug dosages, nutritional levels) and varying effects. An auxiliary-based independence test (AIT) condition is proposed, that can be used to assess the validity of an instrument in an additive nonlinear, non-constant effects model. It is first shown that if the candidate instrument is valid, then the AIT condition holds. Moreover, the implications of the AIT condition are illustrated, and it is demonstrated that, in certain conditions, AIT conditions are necessary and sufficient to detect all invalid instruments. The AIT condition is also extended to incorporate covariates, and a practical algorithm is presented for its implementation. Experimental evaluations using synthetic and real-world data sets demonstrate the effectiveness and robustness of the approach.
