|
| 组合惩罚下联合均值与方差模型的变量选择 |
| Vabiable selection via combined-penalization in joint mean and variance models |
| 投稿时间：2013-04-03 修订日期：2013-08-01 |
| DOI： |
| 中文关键词: 组合惩罚, 联合均值与方差模型, 变量选择, 惩罚极大似然估计 |
| 英文关键词: combined penalization the joint mean and variance model variable selection penalized maximum likelihood |
| 基金项目:大连民族学院自主科研基金项目;国家自然科学基金青年基金项目 |
|
| 摘要点击次数: 1611 |
| 全文下载次数: 0 |
| 中文摘要: |
| 在生产实践和经济领域中, 控制产品质量的方差就能保证产品的合格品数相对稳定, 所以当前人们对联合均值与方差模型的研究倍感兴趣. 实际情况中, 解释变量不相关的情况也几乎是不可能的, 本文提出了一种组合惩罚, 该惩罚充分利用了岭回归能克服解释变量相关性过高对估计效果的影响, 同时也证明了这样的惩罚具有相合性和Oracle性质. 文章使用该组合惩罚对联合均值与方差模型进行变量选择. 最后的随机模拟的结果表明该模型和方法是有效的. |
| 英文摘要: |
| In the production and econometric area, controlling the variance can guarantee the quality of productions relative stability. So people are very interested in joint mean and variance models nowadays. In general, it is uncommon for explanatory variables to be uncorrelated. So we propose a combined-penalization, which is mixed by the SCAD and ridge. It can outperform the SCAD penalty technique when the correlation among predictors is high. In this paper, we use combined-penalization to select variables in mean and variance models. Simulation study show that this model and method are effective. |
|
View Fulltext
查看/发表评论 下载PDF阅读器 |
| 关闭 |