摘要: |
提出可以使用黄金分割法确定Box-Cox变换中的参数,给出了一种计算步骤。提出了通过多搜索几个区间和回归诊断来更大限度地保证所确定出的参数是合用的。并从理论分析和实例检验两方面说明了与一维格点搜索法相比,使用黄金分割法确定Box-Cox变换中的参计算量较小,容易使计算结果达到较高的精确度,且易用性也有所提高。 |
关键词: 线性回归 Box-Cox变换 一维格点搜索法 黄金分割法 |
DOI: |
分类号: |
基金项目: |
|
Golden section search for actualizing Box-Cox transformation |
|
Abstract: |
This paper brings up that golden section search can be used to search for the parameter in Box-Cox transformation and offers a kind of steps in computing it.By searching for the parameter in more sections and using regression diagnosis,the established parameter is ensured to be more suitable.Theoretical analysis and experiments show:comparing with a dimension lattice search,the searching for the parameter in Box-Cox transformation by golden section search has less calculation,and it is easy to get more accurate result and is more applicable. |
Key words: linear regression Box-Cox transformation a dimension lattice search golden section search |