| 罗艳虹,胡良平.适应性回归分析（Ⅳ）———与非适应性回归分析的比较[J].四川精神卫生杂志,2019,32(2):110-115.,Adaptive regression analysis(Ⅳ)——the comparison between the adaptive regression analysis and non - adaptive regression analysis[J].SICHUAN MENTAL HEALTH,2019,32(2):110-115 |
| 适应性回归分析（Ⅳ）———与非适应性回归分析的比较 |
| Adaptive regression analysis(Ⅳ)——the comparison between the adaptive regression analysis and non - adaptive regression analysis |
| |
| DOI：10.11886/j.issn.1007-3256.2019.02.004 |
| 中文关键词: 适应性回归分析 多重线性回归分析 变量变换 自变量筛选 |
| 英文关键词:Adaptive regression analysis Multiple linear regression analysis Variable transformation Independent variable selection |
| 基金项目:国家高技术研究发展计划课题资助(2015AA020102) |
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 【摘要】 本文目的是分析一个已知真实情况的资料,比较适应性回归分析与非适应性回归分析建模的效果.结论如下:当资料中存在与因变量确有关系的自变量时,ADAPTIVEREG过程具有较好的甄别能力;REG过程具有较好的甄别能力,但需要满足一定条件,即采用"前进法"或"逐步法"筛选自变量,同时还需要"假定模型包含截距项".当资料中不存在与因变量确有关系的自变量时,ADAPTIVEREG过程几乎完全失去了甄别能力;REG过程具有较好的甄别能力,但需要满足一定条件,即采用"前进法"筛选自变量,同时还需要"假定模型包含截距项".若研究者基于"基本常识"和"专业知识"确定的自变量都与因变量有关系,对因变量进行Logistic变换,并且,假定回归模型中不含截距项时,会在回归模型中保留非常多的自变量。 |
| 英文摘要: |
| This article compared the effect of adaptive regression analysis with the effect of non - adaptive regression analysis through a known real data set. The conclusions were as below: while the independent variables had a definite relationship with the dependent variable, the ADAPTIVEREG procedure had a great discriminating ability; the REG procedure had a good screening ability,but it must satisfy two conditions, the first one was that the forward selection or the stepwise selection to screen the independent variables would be used, and the second one was that the model would be assumed to contain an intercept term. While the independent variables in the data set didn't have the relationship with the dependent variable, the ADAPTIVEREG procedure almost completely lost its ability to discriminate; the REG procedure had a good screening ability, but it needed satisfy almost the same two conditions mentioned before. If the independent variables determined by researchers based on their common knowledge and the professional one were related to the dependent variable, a large number of independent variables would be retained in the regression model while the dependent variable was transformed by means of the Logistic transformation and it was assumed that there wasn't an intercept term in the regression model. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|