响应变量缺失下条件平均处理效应的$k$近邻核估计

数学物理学报 ›› 2025, Vol. 45 ›› Issue (3): 992-1012.

• • 上一篇

响应变量缺失下条件平均处理效应的$k$近邻核估计

曾华俊^1,²,明瑞星^1,²,苏培娟^1,²,黄绍航^1,²,肖敏^1,^3,^*()

¹浙江工商大学统计数据工程技术与应用协同创新中心杭州 310018
²浙江工商大学统计与数学学院杭州 310018
³浙江工商大学经济运行态势预警与模拟推演实验室杭州 310018

收稿日期:2024-06-18 修回日期:2024-09-13 出版日期:2025-06-26 发布日期:2025-06-20
通讯作者: *肖敏，Email: xiaomin@zjgsu.edu.cn
基金资助:
浙江省社会科学规划课题(22GXSZ001Z);数字+学科建设项目(SZJ2022B004);浙江省教育厅一般项目(Y202353084);浙江省登峰学科(Zhejiang Gongshang University-Statistics);Summit Advancement Disciplines of Zhejiang Province(Zhejiang Gongshang University-Statistics);浙江省省属高校基本科研业务费专项资金(XT202302)

$k$-Nearest Neighbor Kernel Estimation of Conditional Average Treatment Effect with Missing Response Variables

Zeng Huajun^1,²,Ming Ruixing^1,²,Su Peijuan^1,²,Huang Shaohang^1,²,Xiao Min^1,^3,^*()

¹Collaborative Innovation Center of Statistical Data Engineering Technology & Application, Zhejiang Gongshang University, Hangzhou 310018
²School of Statistics and Mathematics, Zhejiang Gongshang University, Hangzhou 310018
³Economic Forecasting and Policy Simulation Laboratory, Zhejiang Gongshang University, Hangzhou 310018

Received:2024-06-18 Revised:2024-09-13 Online:2025-06-26 Published:2025-06-20
Supported by:
Zhejiang Provincial Philosophy and Social Sciences Planning Project(22GXSZ001Z);Digital Science and Engineering Construction Project(SZJ2022B004);Zhejiang Provincial Department of Education General Project(Y202353084);Fundamental Research Funds for the Provincial Universities of Zhejiang(XT202302)

摘要/Abstract

摘要：

基于 Neyman-Rubin 潜在结果框架, 构建 $k$ 近邻核估计量来测度响应变量随机缺失情形下的条件平均处理效应 (conditional average treatment effect, CATE), 旨在评估不同处理方式对个体的影响. 证明了 $k$ 近邻核估计量的几乎完全收敛性和渐近正态性. 数值模拟表明 $k$ 近邻核估计量的表现优良, 利用真实数据进行实证分析, 实证结果显示 $k$ 近邻核估计量具有较小的平均绝对偏差和均方根误差.

关键词: 条件平均处理效应, 随机缺失, $k$ 近邻核估计量, 渐近正态性

Abstract:

Under the Neyman-Rubin potential outcome framework, we construct a $k$-nearest neighbor kernel estimator to measure the conditional average treatment effect in the case of random missing response variables, aiming to evaluate the impact of different treatments on individuals. The paper proves the almost complete convergence and the asymptotic normality of the estimator. The numerical simulation shows that the $k$-nearest neighbor kernel estimator performs well. The real-world data is used for empirical analysis, and the empirical results show that mean absolute error and root mean square error of the $k$-nearest neighbor kernel estimator are smaller.

Key words: conditional average treatment effect, random missing, $k$-nearest neighbor kernel estimator, asymptotic normality

中图分类号:

O212.7

曾华俊, 明瑞星, 苏培娟, 黄绍航, 肖敏. 响应变量缺失下条件平均处理效应的$k$近邻核估计[J]. 数学物理学报, 2025, 45(3): 992-1012.

Zeng Huajun, Ming Ruixing, Su Peijuan, Huang Shaohang, Xiao Min. $k$-Nearest Neighbor Kernel Estimation of Conditional Average Treatment Effect with Missing Response Variables[J]. Acta mathematica scientia,Series A, 2025, 45(3): 992-1012.

图/表 5

表 1

图1

图2

图3

表 2

参考文献 30

[1]	Rosenbaum P R, Rubin D B. The central role of the propensity score in observational studies for causal effects. Biometrika, 1983, 70(1): 41-55
[2]	Rosenbaum P R, Rubin D B. Constructing a control group using multivariate matched sampling models that incorporate the propensity score. Amer Statist, 1985, 39(1): 33-38
[3]	Caron A, Baio G, Manolopoulou I. Estimating individual treatment effects using non-parametric regression models: A review. J Roy Statist Soc Ser A, 2022, 185(3): 1115-1149
[4]	Yao L Y, Chu Z X, Li S, et al. A survey on causal inference. ACM Transactions on Knowledge Discovery from Data, 2021, 15(5): 1-46
[5]	Sarraju A, Ward A, Li J, et al. Personalizing cholesterol treatment recommendations for primary cardiovascular disease prevention. Sci Rep, 2022, 12(1): Article 23
[6]	Sorace M. The ties that unbind: Intergovernmental decision rules and the policy-opinion link. J Eur Public Policy, 2023, 30(8): 1609-1632
[7]	Zhang Z W, Liu W, Zhang B, et al. Causal inference with missing exposure information: Methods and applications to an obstetric study. Stat Methods Med Res, 2016, 25(5): 2053-2066 pmid: 24318273
[8]	Abrevaya J, Hsu Y C, Lieli R P. Estimating conditional average treatment effects. J Bus Econom Statist, 2015, 33(4): 485-505
[9]	Li L, Zhou N W, Zhu L X. Outcome regression-based estimation of conditional average treatment effect. Ann Inst Statist Math, 2022, 74(5): 987-1041
[10]	Kudraszow N L, Vieu P. Uniform consistency of kNN regressors for functional variables. Statist Probab Lett, 2013, 83(8): 1863-1870
[11]	Ling N X, Meng S Y, Vieu P. Uniform consistency rate of kNN regression estimation for functional time series data. J Nonparametr Stat, 2019, 31(2): 451-468
[12]	Friedman J H. Greedy function approximation: A gradient boosting machine. Ann Statist, 2001, 29: 1189-1232
[13]	Chen T Q, Guestrin C. Xgboost: A scalable tree boosting system//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, 785-794
[14]	Breiman L. Random forests. Mach Learn, 2001, 45(1): 5-32
[15]	Wager S, Athey S. Estimation and inference of heterogeneous treatment effects using random forests. J Amer Statist Assoc, 2018, 113(523): 1228-1242
[16]	Athey S, Wager S. Estimating treatment effects with causal forests: An application. Observ Stud, 2019, 5(2): 37-51
[17]	Little R J, D'agostino R, Cohen M L, et al. The prevention and treatment of missing data in clinical trials. New Engl J Med, 2012, 367(14): 1355-1360
[18]	Ocampo A, Schmidli H, Quarg P, et al. Identifying treatment effects using trimmed means when data are missing not at random. Pharm Stat, 2021, 20(6): 1265-1277 doi: 10.1002/pst.2147 pmid: 34169641
[19]	Kuzmanovic M, Hatt T, Feuerriegel S. Estimating conditional average treatment effects with missing treatment information//Proceedings of the 26th International Conference on Artificial Intelligence and Statistics. Valencia: PMLR, 2023, 746-766
[20]	Stensrud M J, Robins J M, Sarvet A, et al. Conditional separable effects. J Amer Statist Assoc, 2023, 118(544): 2671-2683
[21]	Zhao A Q, Peng D. To adjust or not to adjust? Estimating the average treatment effect in randomized experiments with missing covariates. J Amer Statist Assoc, 2024, 119(545): 450-460
[22]	苗旺, 刘春辰, 耿直. 因果推断的统计方法. 中国科学: 数学, 2018, 48(12): 1753-1778
	Miao W, Liu C C, Geng Z. Statistical approaches for causal inference (in Chinese). Sci Sin Math, 2018, 48(12): 1753-1778 (in Chinese).
[23]	Greenland S, Pearl J, Robins J M. Confounding and collapsibility in causal inference. Statist Sci, 1999, 14(1): 29-46
[24]	Geng Z, Guo J H, Fung W K. Criteria for confounders in epidemiological studies. J R Stat Soc Ser B Stat Methodol, 2002, 64(1): 3-15
[25]	Han P S. Multiply robust estimation in regression analysis with missing data. J Amer Statist Assoc, 2014, 109(507): 1159-1173
[26]	Ferraty F, Laksaci A, Tadj A, et al. Rate of uniform consistency for nonparametric estimates with functional variables. J Statist Plann Inference, 2010, 140(2): 335-352
[27]	Hill J L. Bayesian nonparametric modeling for causal inference. J Comput Graph Statist, 2011, 20(1): 217-240
[28]	Burba F, Ferraty F, Vieu P. $k$-Nearest neighbour method in functional nonparametric regression. J Nonparametr Stat, 2009, 21(4): 453-469
[29]	Ezzahrioui M, Ould-Saïd E. Asymptotic normality of a nonparametric estimator of the conditional mode function for functional data. J Nonparametr Stat, 2008, 20(1): 3-18
[30]	黄绍航. 随机缺失下条件平均处理效果的 $k$ 近邻核估计. 杭州: 浙江工商大学, 2022
	Huang S H. $K$-nearest Neighbor Kernel Estimation of Conditional Average Treatment Effect Under Random Missing Data. Hangzhou: Zhejiang Gongshang University, 2022

响应变量缺失下条件平均处理效应的$k$近邻核估计

$k$-Nearest Neighbor Kernel Estimation of Conditional Average Treatment Effect with Missing Response Variables

RichHTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 5

参考文献 30

相关文章 14

Metrics

本文评价

推荐阅读 10

[1]	饶珍敏, 王江峰, 胡康, 何姗. 删失指标随机缺失下一般线性模型的加权最小二乘估计[J]. 数学物理学报, 2025, 45(3): 919-933.
[2]	尹长明,石岳鑫. 纵向数据下自适应设计广义估计方程的大样本性质[J]. 数学物理学报, 2021, 41(6): 1925-1936.
[3]	马奕佳,薛留根,芦飞. 缺失数据下部分非线性变系数EV模型的统计推断[J]. 数学物理学报, 2020, 40(2): 460-474.
[4]	王继霞, 肖庆宪. 非时齐扩散模型中扩散系数的局部估计[J]. 数学物理学报, 2015, 35(2): 332-344.
[5]	陈盼盼, 冯三营, 薛留根. 缺失数据下半参数变系数部分线性模型的统计推断[J]. 数学物理学报, 2015, 35(2): 345-358.
[6]	王成勇. 半参数平滑转换回归模型及其级数估计[J]. 数学物理学报, 2012, 32(4): 797-807.
[7]	冯三营, 薛留根. 非线性半参数EV模型的最大经验似然估计[J]. 数学物理学报, 2012, 32(4): 729-743.
[8]	凌能祥, 丁洁. 相依函数型数据条件密度估计的渐近性质[J]. 数学物理学报, 2012, 32(3): 547-556.
[9]	张娟, 崔恒建. 缺失数据下EV模型的调整最小二乘估计[J]. 数学物理学报, 2009, 29(6): 1465-1476.
[10]	夏天; 唐年胜; 王学仁; 张文专. 非线性再生散度随机效应模型的渐近性质[J]. 数学物理学报, 2007, 27(1): 146-154.
[11]	李永明, 杨善朝. NA样本密度函数估计一致渐近正态性的收敛速度[J]. 数学物理学报, 2005, 25(5): 643-651.
[12]	陈桂景, 陈宇明, 白志东, 胡飞芳. 概率罐子模型中一种最优设计[J]. 数学物理学报, 2003, 23(2): 208-214.
[13]	唐年胜朱仲义韦博成. 非线性再生散度模型的渐近性质[J]. 数学物理学报, 2000, 20(3): 321-328.
[14]	王启华. 随机删失下位置刻度模型中的强相合与渐近正态估计[J]. 数学物理学报, 1999, 19(1): 1-9.