“Unlike the existing parametric approaches, our method is simple yet flexible to encapsulate distributional dependence structures of bivariate outcomes and covariates. Various simulation results confirm that our method can perform similarly or better in finite samples compared to the alternative methods.”