-
首页
-
- 首页
- WORKSHOP预告
劳动与健康经济学workshop: Exact variance of average treatment effect estimator in cluster RCT
发布日期：2023-09-21 12:00 来源：
时间：2023年9月21日（周四）10:00-11:30
地点：北京大学国家发展研究院承泽园院区 246教室
主持人：雷晓燕、易君健
参与老师：赵耀辉、李玲、刘国恩、张丹丹、黄炜
主讲人：方悦 香港中文大学（深圳）
摘要：In this paper, we discuss the average treatment effect estimator in clustered randomized trials with the finite population setting. Instead of focusing on the treatment effect estimator itself, we consider how to find a better variance estimator of the corresponding ATE estimator. In clustered randomized trials, units of the same cluster are assigned to the same treatment group, either treated or control. Therefore, either the treated or the control outcome can be observed for each cluster. We consider the setting of sampling and treatment assignment without replacement, and calculate the exact variance of the ATE estimator. There is a term on the covariance between the cluster average potential outcome when treated and cluster average potential outcome when controlled, which cannot be estimated directly. Liang-Zeger estimator, which is a benchmark and is widely adopted in applied microeconomics, omits this term and is usually too conservative. By referring to Aronow et al. (2014) for the bound on the covariance term, we propose a new variance estimator in this setting, which could improve over Liang-Zeger estimator. We illustrate the performance of our estimator using both simulation and application results.
主讲人介绍：方悦（Yue Fang），香港中文大学（深圳）经管学院助理教授，2023年于南加州大学取得经济学博士学位，此前2017年于北京大学取得经济学学士学位。研究领域为计量经济学和实证微观经济学，研究兴趣包括因果推断、政策学习和机器学习等。
国家发展研究院官方微信
Copyright© 1994-2012 北京大学 国家发展研究院 版权所有, 京ICP备05065075号-1
保留所有权利，不经允许请勿挪用
-
首页
-
劳动与健康经济学workshop: Exact variance of average treatment effect estimator in cluster RCT
发布日期：2023-09-21 12:00 来源：
时间：2023年9月21日（周四）10:00-11:30
地点：北京大学国家发展研究院承泽园院区 246教室
主持人：雷晓燕、易君健
参与老师：赵耀辉、李玲、刘国恩、张丹丹、黄炜
主讲人：方悦 香港中文大学（深圳）
摘要：In this paper, we discuss the average treatment effect estimator in clustered randomized trials with the finite population setting. Instead of focusing on the treatment effect estimator itself, we consider how to find a better variance estimator of the corresponding ATE estimator. In clustered randomized trials, units of the same cluster are assigned to the same treatment group, either treated or control. Therefore, either the treated or the control outcome can be observed for each cluster. We consider the setting of sampling and treatment assignment without replacement, and calculate the exact variance of the ATE estimator. There is a term on the covariance between the cluster average potential outcome when treated and cluster average potential outcome when controlled, which cannot be estimated directly. Liang-Zeger estimator, which is a benchmark and is widely adopted in applied microeconomics, omits this term and is usually too conservative. By referring to Aronow et al. (2014) for the bound on the covariance term, we propose a new variance estimator in this setting, which could improve over Liang-Zeger estimator. We illustrate the performance of our estimator using both simulation and application results.
主讲人介绍：方悦（Yue Fang），香港中文大学（深圳）经管学院助理教授，2023年于南加州大学取得经济学博士学位，此前2017年于北京大学取得经济学学士学位。研究领域为计量经济学和实证微观经济学，研究兴趣包括因果推断、政策学习和机器学习等。
访问学者申请
EMAIL
北京大学
ENGLISH
