Many Average Partial Effects: with An Application to Text Regression

Abstract

We study estimation, pointwise and simultaneous inference, and confidence intervals for many average partial effects of lasso Logit. Focusing on high-dimensional, cluster-sampling environments, we propose a new average partial effect estimator and explore its asymptotic properties. Practical penalty choices compatible with our asymptotic theory are also provided. The proposed estimator allow for valid inference without requiring oracle property. We provide easy-to-implement algorithms for cluster-robust high-dimensional hypothesis testing and construction of simultaneously valid confidence intervals using a multiplier cluster bootstrap. We apply the proposed algorithms to the text regression model of Wu (2018) to examine the presence of gendered language on the internet.

References in corpus (5)

• Confidence Intervals and Hypothesis Testing for High-Dimensional Regression
• Robust Inference on Average Treatment Effects with Possibly More Covariates than Observations
• On the Distribution of Penalized Maximum Likelihood Estimators: The LASSO, SCAD, and Thresholding
• Debiased Inference of Average Partial Effects in Single-Index Models
• Delta Theorem in the Age of High Dimensions