回归系数在回归方程中表示自变量x 对因变量y 影响大小的参数。 回归系数越大表示x 对y 影响越大，正回归系数表示y 随x 增大而增大，负回归系数表示y 随x 增大而减小。 　　回归方程式^Y=bX+a中之斜率b,称为回归系数,表X每变动一单位,平均而言，Y将变动b单位。

标准回归系数消除了因变量y和自变量x1，x2，…xn所取单位的影响之后的回归系数，其绝对值的大小直接反映了xi对y的影响程度。

计算方法

　　对数据标准化，即将原始数据减去相应变量的均数后再除以该变量的标准差，计算得到的回归方程称为标准化回归方程，相应得回归系数为标准化回归系数。

原链接：http://iask.sina.com.cn/b/17115841.html?sudaref=www.google.com.hk&retcode=0

您好，我想问一下，如果变量都是为比例变量，有必要标准化吗？
还有其经济意义如何？举例说明，回归方程如果是这样的：
y=c+b1x1+b2x2+b3X3,X,Y都是百分比，比较x1,x2,x3对y的重要性需要标准化回归吗？
回归方程如果是这样的：y=c+b1x1+b2x2+b3X3+lnx4,变量x4需以对数形式放进方程，其他变量还是百分比，这时比较x1,x2,x3对y的重要性需要标准化回归吗？标准化后lnx4的经济意义有无变化？

实证中经常出现的自变量变化一个标准差因变量变化多少是怎么算出来的？

人大论坛：http://bbs.pinggu.org/thread-513629-1-1.html

某人的Stata 博客，里面很多帮助信息

Table of content

• How can I make the example work?
• Why don't you just give the output?
• What does sysuse auto, clear mean?
• Why do you put things like long, double, or str8 between variable names when you use input?
• Why do you recode rep78 1/2 = 3?
• Why do I add the variable baseline together with the nocons option?
• What does gen rep3 = rep78 == 3 if rep78 < . do?
• What does gen rep3Xprice = (rep78 == 3)*price if rep78 < . do?
• Why do you sometimes add if rep78 < . to a command?
• What does !missing(var1,var2) mean?
• What does gen domestic = !foreign do?
• What does gen marst = !never_married + married if !missing(never_married, married) do?

Re: st: margins for 1 standard deviation change

I need to get the discrete (1 standard deviation, not 0->1) change for

binary variables in my logit equation, just like:

clear

webuse auto

recode rep78(0/3 = 1) (3/99 = 0)

logit for price mpg head rep78

mfx, at(mean)

logit for price mpg head i.rep78

margins, dydx(*) atmeans

but this shows only 0->1 discrete change for my new 'rep78' variable.

 How can I output 1 stddev change(or 1/2 std dev change) in margins??

(or mfx, though I prefer to use margins since it understands factor

variables)

*

*   For searches and help try:

--- On Wed, 12/1/11, David Souther wrote:

> I need to get the discrete (1 standard deviation,

> not 0->1) change for binary variables in my logit

> equation, How can I output 1 stddev change(or 1/2

> std dev change) in margins??

*------------------ begin example ----------------

sysuse auto, clear

recode rep78(0/3 = 1) (3/max = 0)

sum rep78 if ///

 !missing(rep78, foreign, price, mpg, headroom)

gen crep = (rep78 - r(mean))/r(sd)

logit foreign price mpg headroom crep

margins, dydx(*) atmeans

*------------------ end example -------------------

(For more on examples I sent to the Statalist see: 

http://www.maartenbuis.nl/example_faq )

Hope this helps,

Maarten

--------------------------

Maarten L. Buis

Institut fuer Soziologie

Universitaet Tuebingen

Wilhelmstrasse 36

72074 Tuebingen

Germany

http://www.maartenbuis.nl