Premium
Optimal forecast intervals under asymmetric loss
Author(s) -
Demetrescu Matei
Publication year - 2007
Publication title -
journal of forecasting
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.543
H-Index - 59
eISSN - 1099-131X
pISSN - 0277-6693
DOI - 10.1002/for.1019
Subject(s) - interval (graph theory) , realization (probability) , context (archaeology) , extension (predicate logic) , econometrics , estimation , prediction interval , function (biology) , computer science , mathematics , mathematical optimization , statistics , economics , paleontology , management , combinatorics , evolutionary biology , biology , programming language
An optimality criterion for forecast intervals under asymmetric loss functions is proposed. A loss optimal forecast interval is obtained by requiring that the expected loss, conditional on a future realization within the desired interval, be minimal. The main difficulty in the context of forecasting under asymmetric loss emerges when there is no knowledge about the distribution of the innovations. For solving this problem, an extension of estimation under the relevant loss function is suggested. In many cases, one also needs to account for the additional variability due to estimation of model parameters. Another solution, based on the bootstrap, works for both problems. Copyright © 2007 John Wiley & Sons, Ltd.