posis: Command for the sure-independence-screening Neyman orthogonal estimator

Inference for structural parameters in a high-dimensional model has become increasingly popular. Belloni, Chernozhukov, and Wei (2016, Journal of Business and Economic Statistics 34: 606–619) proposed a lasso-based Neyman orthogonal estimator that produces valid inference for the coefficients of interest in the generalized linear model. Drukker and Liu (2022, Econometric Reviews 41: 1047–1076) extend their estimator by using a Bayesian information criterion (BIC) stepwise-based Neyman orthogonal estimator, and the simulations show the advantage of using BIC-based stepwise as the covariate-selection technique. However, the BIC-stepwise-based Neyman orthogonal estimator becomes compu- tationally infeasible when there are many more control variables. To overcome this computational bottleneck, Drukker and Liu (2022) proposed combining the sure- independence-screening technique with BIC-based stepwise to improve the compu- tational speed while maintaining similar or better statistical performance. In this article, we present posis, a command for an iterative-sure-independence-screening- based Neyman orthogonal estimator for the high-dimensional linear, logit, and Poisson models.