Anthony Obrzut
Title: Addressing Regressor Endogeneity via Copulas using R package Rcope.
Date: Thursday, October 29th, 2026
Time: 2:30pm
Location: ASB 10920 / Zoom
Supervised by: Hui Xie
Abstract: Regressor endogeneity poses a significant challenge when drawing causal inferences from observational data. Although instrumental variables (IVs) area classical solution to endogeneity, finding a valid instrument is often difficult. Thus, researchers are investigating copulas as an alternative because of their ability to model the dependence between the endogenous regressor and structural error. Copula-based methods exploit this dependence structure to construct control functions, which are included as generated regressors to account for endogeneity. One such copula-based method is two-stage copula endogeneity-correction (2sCOPE) (Yang, Qian, and Xie, 2025), a computationally tractable approach that does not require IVs. Current copula-based endogeneity correction methods require the endogenous regressor to be non -normal and uncorrelated with the exogenous regressor. 2sCOPE relaxes these requirements, resulting in a more generalizable model. This thesis introduces Rcope, an R package implementing 2sCOPE with functionality for model estimation and endogeneity testing. We demonstrate its usage in a real-world application involving diaper sales data and compare its finite-sample performance with existing endogeneity correction software implementations, showing that Rcope provides a viable alternative when valid IVs are unavailable.