经济计量建模不必始于一般模型的一个注记 阅读全文
下载全文 |
Title | A Note on not Beginning With a General Model
|
作者 | 魏瑾瑞 朱建平 谢邦昌 |
Author | Wei Jinrui, Zhu Jianping and Xie Bangchang |
作者单位 | 厦门大学经济学院统计系,厦门大学数据挖掘研究中心 |
Organization | School of Economics, Xiamen University, Fujian |
作者Email | wejerry@139.com |
中文关键词 | 增元回归 减元回归 正交 相关 偏回归 |
Key Words | Increase Variable Regression; Reduce Variable Regression; Orthogonal; Related; Partial Regression |
内容提要 | 以一般的线性回归模型为例,假定解释变量与扰动项不相关,论证了如果解释变量彼此正交,那么回归模型不必始于一般形式;如果解释变量存在一定程度的相关关系,那么虽然解释变量的系数会因增减其他解释变量而改变,但是解释变量的系数不会因相关性的存在而失去偏回归的意义。对于“遗漏解释变量”问题,从理论假定或经验判断的往往是该变量是否重要,而统计推断则更关心的是该变量与现存解释变量相关与否。 |
Abstract | In general linear regression model for example, on the condition of the independence between explanatory variables and disturbance: if explanatory variables are orthogonal, the regression model should not begin with a general one; if there exist correlation between the explanatory variables, the general model is always the first choice. Secondly, although the coefficient of explanatory variables would change while increasing or decreasing explanatory variables if they are correlated, the meaning of partial regression is not destroyed. the omitted explanatory variables, important or not,is concerned in theory;While statistical inference focuses on the correlation. |
文章编号 | WP122 |
登载时间 | 2011-09-16 |
|