5342245 发表于 2010-3-12 20:29 
我的数据自变间存在严重的共线性,且自变量个数远大于观察次数,现在需建立高维自变量与多个因变量的关系,不知什么统计方法适用?
请大家指点~
You may take a look of a shrinkage estimation class. Shrinkage can be thought of as “constrained” minimization or penalized regression. The idea is to trade of bias and variance.
Here is a little bit more about it.
http://www.pinggu.org/bbs/viewthread.php?tid=681147&page=1#pid4570814