Deep partial least squares for instrumental variable regression

In this paper, we propose deep partial least squares for the estimation of high-dimensional nonlinear instrumental variable regression. As a precursor to flexible neural network architecture, our methodology uses dimension reduction and feature selection from set instruments covariates. A central theoretical result, due Brillinger (2012) Selected Works Daving Brillinger. 589-606, shows that provided by is consistent weights are estimated up proportionality constant. We illustrate with synthetic datasets sparse correlated structure draw applications effect childbearing on mother's labor supply based classic data Chernozhukov et al. Ann Rev Econ. (2015b):649–688. The results as well show method significantly outperforms other related methods. Finally, conclude directions future research.