bangmingshaw 发表于 2010-10-7 21:07 
在设定误差这一讲中,如果遗漏的变量与模型中的变量是相关的,则参数估计的结果会是有偏的,但是我们在处理多重共线性问题时,把相关的变量剔除了,这样估计的结果不都是有偏的了吗?本人菜鸟,请高手指教……
When there exists highly correlated covariates (not perfectly correlated), your claims is absolutely right. So sometimes you have to "balance" your choices under your objectives, for example unbiasedness and predictiveness.