| 公告主旨 |
2026 年 11 月 26 日 CRETA Seminar - Robust estimation and inference with given marginals (G. Franguridi, J. Hahn, A. Kapteyn, G. Ridder)
|
| 公告內容 |
【CRETA Seminar】
時間:2026 年 11 月 26 日 (週四) 下午 1:30~3:00
地點:國立臺灣大學管理學院二號館 3 樓 304 教室
講者:Prof. Grigory Franguridi (Center for Economic and Social Research, University of Southern California) https://www.franguridi.com/
演講主題:Robust estimation and inference with given marginals (G. Franguridi, J. Hahn, A. Kapteyn, G. Ridder)
講題摘要:We consider two-step estimation of a finite-dimensional target parameter, where the first step is the nonparametric estimation of a joint density given known marginals, and the second step is the moment estimation of the target parameter. We derive the influence function for this estimation procedure and show that it depends on a nonparametric additive regression. We explain how to compute the influence function using a version of the extensively studied smooth backfitting algorithm. This leads to a Neyman-orthogonal (debiased) estimator for the target parameter and a procedure for asymptotically valid inference. We consider two applications of our results in detail: a panel data model with (nonparametric) nonignorable attrition and refreshment samples (Hirano et al., 2001) and the optimal transport with entropic regularization (Cuturi, 2013; Galichon and Salanié, 2022). We show in Monte Carlo simulations that our estimator indeed reduced bias, and the associated confidence intervals exhibit coverage close to nominal.
欲參加 CRETA Seminar 的朋友們,煩請事先報名。 報名期限:2026/11/25 (三) 中午 12:00 歡迎踴躍報名參加!
|