二元逻辑回归模型中的修正Liu估计
Modified Liu Estimator in Binary Logistic Regression Model
肖松 1黄介武1
作者信息
- 1. 贵州民族大学 数据科学与信息工程学院,贵州 贵阳 550025
- 折叠
摘要
在二元逻辑回归模型中,基于修正岭型估计和Liu估计的思想,提出了一类新的有偏估计作为极大似然估计的替代估计,以期克服多重共线性的影响.在均方误差矩阵意义下,探讨得到了修正 Liu 估计优于 Liu估计、修正岭型估计、极大似然估计等的条件,通过数值模拟和实证分析对部分理论结果在均方误差准则下进行了验证.结果表明,在一定条件下,当模型存在多重共线性时,修正Liu估计优于Liu估计、修正岭型估计以及极大似然估计.
Abstract
In a binary logistic regression model,based on the modified ridge-type estimator and Liu estimator,a new class of biased estimators is proposed as an alternative to maximum likelihood estimator in order to overcome the influence of multicollinearity.In the meaning of mean square error matrix,the conditions that the modified Liu estimator is better than the Liu estimator,the modified ridge-type estimator and the maximum likelihood estimator are discussed.Some theoretical results are verified under the mean square error criterion by numerical simulation and empirical analysis.The results show that in certain conditions,when the model has multicollinearity,the modified Liu estimator is better than the Liu estimator,the modified ridge-type estimator and the maximum likelihood estimator.
关键词
多重共线性/二元逻辑回归模型/修正Liu估计/均方误差准则Key words
multicollinearity/binary logistic regression model/modified Liu estimator/mean square error criterion引用本文复制引用
基金项目
贵州省高等学校大数据分析与智能计算重点实验室(黔教技[2023]012号)
贵州省高等学校智能算法与智能软件协同创新团队(黔教技[2023]061号)
出版年
2024
NETL
NSTL
维普
万方数据