in quantreg the err or distant in objective function is measured as abs(y--x*b) (L1 problem) while in usual regression the err is measured as (y--x*b)**2 (L2 problem).
So importance of each data point will depends the distance in L1 and distance^2 in L2.
So the estimates in L2 is much easier impacted by a 'bad'/away from average point than that in L1. But it has a close form solution and easy to deal with in math and in computation.
That is why quantreg is said as robust regression.
发帖
雷达卡
加好友,备注cda
京公网安备 11010802022788号
论坛法律顾问：王进律师
知识产权保护声明
免责及隐私声明
