Comment.

In this article authors comment on the paper entitled "Minimax Aspects of Bounded-Influence Regression," by Peter J. Huber. Authors remark that they would like to thanks Huber for a very stimulating analysis of their results. They were surprised by Huber's remark, both at the end and beginning of the research paper, that bounded-influence regression is not concerned with gross errors in the explanatory variables. According to them Huber's set up is not so symmetric. His minimax framework leads to a specific gross-error distribution for y, but assumes that the x's are error-free. Huber's main thesis is that leverage groups should be treated differently from individual leverage points, because it is unlikely that nature would concentrates the gross errors on a particular region of the x space. Professor Huber's theoretical results apply only to one parameter, through the-origin regression, and it is not clear how they would extend to multivariate regression. The author also presents their ideas for the solution of the problems raised by Huber in his paper.