摘要
本文利用SAS帮助数据库中的一个数据集sashelp.enso,介绍对自变量进行样条变换后的曲线回归分析方法。在SAS/STAT的TRANSREG过程中,涉及到六种样条变换方法,分别为:B-样条变换、B-样条基函数变换、单调B-样条变换、非迭代惩罚B-样条变换、迭代光滑样条变换、非迭代光滑样条变换。获得的结论是:在确保 R 2≈0.7且回归模型尽可能精简的条件下,“非迭代惩罚B-样条变换”与“迭代光滑样条变换”两种方法是以上六种方法中最好的曲线回归建模方法,这两种方法的拟合效果几乎完全相同。
This paper was to introduce the approaches of curve regression analysis through the spline transformation of the independent variable by means of using the data set named sashelp.enso in the data base of SAS HELP. In the TRANSREG procedure of the SAS/STAT, six approaches of the spline transformation were involved as below: B-spline transformation , B-spline base transformation , monotonic B-spline transformation, non-iterative penalized B-spline transformation, iterative smoothing spline transformation , non-iterative smoothing spline transformation. The conclusion were as follows: under the conditions of ensuring the R-square to be equal to 0.7 approximately and the regression model streamlining as much as possible, the fourth and fifth approach mentioned above were the best and they had almost the same fitting effects.
作者
胡良平
Hu Liangping(Graduate School, Academy of Military Sciences PLA China, Beijing 100850, China;Specialty Committee of Clinical Scientific Research Statistics of World Federation of Chinese Medicine Societies, Beijing 100029, China)
出处
《四川精神卫生》
2019年第3期197-202,共6页
Sichuan Mental Health
基金
国家高技术研究发展计划课题资助(2015AA020102)
关键词
曲线回归
非迭代惩罚B-样条变换
光滑样条变换
节点
光滑参数
Curve regression
Noniterative penalized B-spline transformation
Smoothing spline transformation
Knot
Smoothing parameter
