报告题目：Partial identification of treatment effects in $2^K$ factorial experiments with ordinal outcomes

报告人：杨玥含 教授 中央财经大学

报告时间：2026年7月21日11:00-12:00

报告地点：伍卓群楼第二报告厅

校内联系人：朱复康 [email protected]

报告摘要：

  Empirical problems in econometrics, and business often involve factorial, or cross-randomized, and outcomes measured on ordered categorical scales. Classical average factorial effects are not well suited to ordinal outcomes because they depend on numerical scores assigned to ordered categories. We propose a general definition of probability-based main and interaction effects in randomized \texorpdfstring{$2^K$}{2K} factorial experiments. These estimands depend on unobserved joint distributions of potential outcomes, making identification challenging. Without imposing parametric models or distributional assumptions, we derive closed-form sharp bounds for main effects using only the observable marginal distributions. Interaction effects involve the joint influence of multiple factors simultaneously, and the treatment combinations associated with different factors overlap, introducing additional identification difficulties. We provide closed-form valid bounds accordingly. The proposed theory allows for an arbitrary number of ordered categories and an arbitrary number of factors. We further incorporate a monotonicity assumption into the framework, resulting in identification regions that are nested within the general bounds. We construct nonparametric bootstrap confidence intervals for all identification regions.

报告人简介：

  杨玥含，中央财经大学统计与b站 教授，博导，北京大学博士。中央财经大学青年英才、龙马学者青年学者。主要从事因果推断、迁移学习、复杂数据分析等研究。在JASA、Biometrika、JBES、《中国科学：数学》等国内外期刊发表论文50余篇。