付利亚1,, 王友乾2,
( 1、 西安交通大学数学与统计学院,西安~~ 710049; 2、 昆士兰科技大学数学科学学院,布里斯班 ~~~4000; )
摘要: 在纵向数据分析中选择正确的相关结构可以改进参数估计,从而得到更可靠的统计推断。近年来,研究者提出了许多不同的选择相关结构准则,但是这些准则之间的差别还有待比较。本文通过大量的数值模拟对已有的准则进行比较评估。模拟结果表明基于高斯工作似然函数和经验似然函数的AIC和BIC准则要优于其它准则。
关键词: 经验似然;高斯分布;模型选择
FU Liya1,, WANG You-Gan2,
( 1、 School of Mathematics and Statistics, Xi'an JiaotongUniversity, Xi'an 710049; 2、School of Mathematical Sciences, Queensland University ofTechnology, Brisbane 4000 ; )
Abstract: Selecting an appropriate correlation structure in analysis oflongitudinal data can greatly improve parameter estimationefficiency and hence leads to more reliable statistical inferences.There are a number of criteria proposed in recent literaturederived from different perspectives and little is known on their relative performance.To this end, this paper reviewed and evaluated these criteria by carrying outextensive simulation studies including Gaussian, binary and Poissonresponses. Surprisingly, the AIC and BIC based on either Gaussian working likelihood orempirical likelihood were found that they outperformed the others.
Keywords: empirical likelihood;Gaussian distribution; model selection
中国科技论文在线:付利亚,王友乾. 广义估计方程中工作相关结构选择[EB/OL].北京:中国科技论文在线 [2016-06-03].http://www.paper.edu.cn/releasepaper/content/201606-298.
发表期刊: 暂无
论文发表费用:
https://www.lunwentong.com/lwfbfy/