edragon1983 发表于 2010-1-9 21:52 
本人不是学数理统计的，在看文献时，遇到了lasso logistic regression。请问我们常用的logistic regression 就是lasso法吗？
如果不是，两者有啥区别和联系?
谢谢！
Lasso regression is closely related to Ridge regression. Both belong to the shrinkage estimation class.
Here is the motivations to introduce Lasso (Ridge) regression. When any two of predictors is highly correlated to each other, the det(x'x) is close to zero. The result is larger estimation variance. The Lasso and Ridge regression is to trade off between unbiasness and a smaller estimation variance by putting a penalty factor/constrain on coeffecients betas. Different type of constrains(absolute / square) will result in Lasso/Ridge regression.
Any advanced econometrics/statistics testbook should cover the topic.
扫码加好友，拉您进群
全部版块
我的主页
收藏
