摘要
利用复合分位数回归(CQR)研究了纵向数据下回归系数的经验似然(EL)推断。将数据组内相关性纳入估计方程中,得到了参数更加有效和稳健的估计,无需预先估计渐近方差,从理论上证明了所得复合分位数经验似然估计(CQREL)的渐近性质,且提出的经验对数似然比服从渐近卡方分布,从而构造了兴趣参数的置信域。通过数值模拟验证了所得方法有很好的有限样本性质。
In this paper,empirical likelihood-based inference procedure via composite quantile regression with longitudinal data is investigated.By incorporating correlation within subjects,the proposed method produces more efficient and robust estimators.The theoretical properties of the resulting composite quantile regression empirical likelihood estimator(CQREL)are established.The proposed empirical log-likelihood ratio statistics has an asymptotic standard chi-square distribution,and hence the confidence region for the parameter of interest can be constructed.Simulation studies indicate that the proposed empirical likelihood procedure has good performance.
作者
刘佳乐
黄彬
LIU JiaLe;HUANG Bin(College of Mathematics and Physics,Beijing University of Chemical Technology,Beijing 100029,China)
出处
《北京化工大学学报（自然科学版）》
CAS
CSCD
北大核心
2020年第1期107-112,共6页
Journal of Beijing University of Chemical Technology(Natural Science Edition)
基金
国家自然科学基金(11471321)。
