M-estimator is a BIG class which depends on how an objective functional form is defined. L-estimator is one of them.
Here is the link for M-estimators.

http://en.wikipedia.org/wiki/M-estimator

ols不是M类估计,
M估计是在样本中有异常值的时候为了使方程中系数受异常值影响小一些

错!  错!

What is M estimator, there is a clear definition. You can see it either in Woodridge's or Greene's book. Or the link I give in my post.

M estimator is a big class. Some of them are robust estimatiors in the sense THEY are insensitive to a 异常值.

Don't comfuse M estimator with ROBUST estimatior.

Hope this helps.

2# bobguy
THANKS.学习了,收获!

之所以出现这个困惑,在于CAMERON 在'"Microeconometrics : Methods and Applications"一书中关于M-类估计的一个注释,原文如下:
The term m-estimator, attributed to Huber (1967), is interpreted as an abbreviation
for maximum-likelihood-like estimator. Many econometrics authors, including Amemiya (1985, p. 105), Greene (2003, p. 461), andWooldridge (2002, p. 344), define an m-estimator as optimizing over a sum of terms, as in (5.4). Other authors, including Serfling (1980), define an m-estimator as solutions of  quations such as (5.5). Huber (1967) considered both cases and Huber (1981, p. 43) explicitly defined an m-estimator in both ways. In this book we call the former type of estimator an m-estimator and the latter an estimating equations estimator (which will be treated separately in Section 5.4).
知道ML 与 NLS均是M-类估计的典型代表.但对于OLS是不是M-类估计,与同学发生了分歧.根据2# bobguy 的说明,现下已经有了判断!再谢!