ht此去经年 发表于 2011-6-18 21:19 
求助！计量经济学中，对不同单位的变量建模研究时，数据处理取对数后是否需要再对其进行标准化处理？取对数能否消除量纲影响？
It depents a model assumption. In a classical regression model the error term is additive and model is linear form, example
y=a + b*x +err
In a log model, we usually assume the error term is multiplicative, the original model would be
y=a*x^b*err
log(y)=log(a)+b*log(x)+log(err)
Redefine variables and assume that err is lognormal, then the equation is a classical regression as,
y2 = a2 + b*x2 + err2. where y2=log(y) , a2=log(a), x2=log(x), log(err)= Normal distribution.
扫码加好友，拉您进群
全部版块
我的主页
收藏
