jiangshq
以前经常用SAS编程分析数据,进来迷上R,在SAS中逐步回归很容易实现直接加stepwise就可以了,在R中本人尚没有找到合适的函数,请高手指点!致谢!
yihui
?step
drewlee
[quote]引用第1楼谢益辉于2007-04-27 13:36发表的“”:
?step[/quote]
R里面的step是以AIC为criteria,而SAS和SPSS都是p-value。
我在寻找R里面能以p-value为criteria的stepwise function。
longoR
leaps
regsubsets
drewlee
[quote]引用第3楼longoR++于2007-04-27 15:40发表的“”:
leaps
regsubsets[/quote]
Are you sure they are stepwise based on p-value?
rtist
based on r2 is equivalent to p-values for linear models
drewlee
[quote]引用第5楼rtist于2007-04-28 00:18发表的“”:
based on r2 is equivalent to p-values for linear models[/quote]
It's shocking to me.
What kind of equivanlence? The selection result? or asymptotic equivanlence?
rtist
given fixed df, they are one-to-one fctns of each other
drewlee
I am confused still at this explanation...
yihui
你说的P-value是单个系数的P-value还是整个模型F统计量的P-value?
drewlee
[quote]引用第9楼谢益辉于2007-04-29 00:36发表的“”:
你说的P-value是单个系数的P-value还是整个模型F统计量的P-value?[/quote]
单个的,SAS里应该是单个系数的P-value。