ols的时候，如果只是 control variables 不是正态分布，还有必要调整至正态分布吗？

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

如题。。谢谢！！

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：Variables Variable control Contro contr OLS 正态分布 control Variables 至正

control variables 是解释变量，应该是确定性变量而不是随机变量

sqy 发表于 2010-2-21 21:13 control variables 是解释变量，应该是确定性变量而不是随机变量

可否把OLSE的假设写出来？

jiliangamelia 发表于 2010-2-21 06:17
如题。。谢谢！！

Why the control variables need to be normally distributed?

What if a control variable is gender(male/female)? How can you normalized the variable.

OLS doesn't require any distributional assumptions about the data, as u might see from the formula.
However for the inference purpose, it might be the way. Plot your control variable vs. response variable first, see whether it suggests linear relationship or not.
If responses are  increasing with the increase in explanatory variable, but the shape is not exactly "straight",  then you might consider transformation.

Also what is the type of your control variable? continues? categorical?

windlove 发表于 2010-2-21 23:04
OLS doesn't require any distributional assumptions about the data, as u might see from the formula.
However for the inference purpose, it might be the way. Plot your control variable vs. response variable first, see whether it suggests linear relationship or not.
If responses are  increasing with the increase in explanatory variable, but the shape is not exactly "straight",  then you might consider transformation.

Also what is the type of your control variable? continues? categorical?

一共11个控制变量，都是连续型的数据。response variables是控制变量的意思吧？

I think you better to find a reference book, coz you might have no background in statistics.

I recommend "Applied Linear Regression 3rd" by SANFORD WEISBERG.  I am sure the book can be d\l from the forum here.

Response variable is a designed outcome variable or a variable you might be of interest by controlling other factors (control variables).  ie. Y = b0+x1*b1+x2*b2+ .....

windlove 发表于 2010-2-22 00:21
I think you better to find a reference book, coz you might have no background in statistics.

I recommend "Applied Linear Regression 3rd" by SANFORD WEISBERG.  I am sure the book can be d\l from the forum here.

下载到了，谢谢推荐！

jiliangamelia 发表于 2010-2-21 23:44

windlove 发表于 2010-2-21 23:04
OLS doesn't require any distributional assumptions about the data, as u might see from the formula.
However for the inference purpose, it might be the way. Plot your control variable vs. response variable first, see whether it suggests linear relationship or not.
If responses are  increasing with the increase in explanatory variable, but the shape is not exactly "straight",  then you might consider transformation.

Also what is the type of your control variable? continues? categorical?

一共11个控制变量，都是连续型的数据。response variables是控制变量的意思吧？

I think you had better to let the professional one do it.

Statistic analysis  is not the job just for anyone who just knows little about almost everything.

Statistic analysis is the combination knowledge of business, data, statistics. Lack anyone of them would be useless(GIGO).