Technical Reports


LaMAsE: Least Asymmetric Mean Absolute Scaled Error Estimation

Wang, Jianqiang.C ; Guler, Kemal
HP Laboratories


Keyword(s): Asymmetric loss; quantile regression; forecast evaluation

Abstract: We consider Asymmetric Mean Absolute Scaled Error (aMAsE), a parameterized extension of scaled absolute error, that penalizes negative and positive errors asymmetrically. We study the theoretical properties of estimates when this measure is used as the loss criterion to be minimized.

31 Pages

External Posting Date: December 8, 2011 [Abstract]. Approved for External Publication
Internal Posting Date: November 8, 2011 [Fulltext]

Back to Index