Inference for Heterogeneous Effects using Low-Rank Estimation of Factor Slopes
研究面板数据中个体和时间异质性效应的推断问题，通过低秩正则化回归估计因子结构斜率，并利用样本分割和偏出方法实现有效推断，适用于个体-时间特定效应及其截面平均的统计推断。
Abstract We study a panel data model with heterogeneous effects, allowing slopes to vary across individuals and time. To reduce dimensionality, we assume these slopes follow a factor structure, so slope matrices can be estimated via low-rank regularized regression. We propose a multi-step estimation procedure incorporating sample splitting and partialing-out to enable valid inference after penalized estimation. We establish the asymptotic normality of the resulting estimator, facilitating inference for individualtime- specific effects and their cross-sectional averages. The method’s performance is illustrated through simulations and an empirical application.