xiongxi 回复 第1楼 的 wchg1268： 你可以google一下rank regression, 其次，我记得SAS中的PLS procedure是可以做减秩回归的，当然也有比较2B青年的做法，首先rank procedure，然后再reg procedure，我昨天在公司就是这么做的（我记得你是要问SAS实现的[s:11]）
micro@ Never heard of this before... Found a pretty clear doc: http://www.math.wustl.edu/~sawyer/handouts/RankRegress.pdf
micro@ Because of the convexness of the objective function, writing a simple implementation should be trivial in R.
micro@ I wonder why such a simple method did not receive much notice in the history (say, compared to M-estimates, least trimmed squares, winsorizing...)? Are there any serious problems with it[s:16] Or, is it just due to my ignorance?
