Premium
Entropy‐based benchmarking methods
Author(s) -
Temursho Umed
Publication year - 2018
Publication title -
statistica neerlandica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.52
H-Index - 39
eISSN - 1467-9574
pISSN - 0039-0402
DOI - 10.1111/stan.12160
Subject(s) - benchmarking , entropy (arrow of time) , computer science , sign (mathematics) , competitor analysis , series (stratigraphy) , artificial intelligence , data mining , mathematics , economics , physics , mathematical analysis , paleontology , management , quantum mechanics , biology
This paper argues that benchmarking of sign‐volatile time series has to be based on the principle of movement and sign preservation , which requires the benchmarked series to reproduce both the short‐term movements and signs of the original series. It is shown that the widely used variants of Denton method and the Causey‐Trager growth rate preservation method may violate this principle. Four variants of entropy‐based benchmarking methods are proposed that, by construction, account for the requirements of the movement and sign preservation principle. Our extensive simulations confirm that the proposed entropy methods can be regarded as plausible competitors for the well‐established benchmarking methods and may even be preferred in cases of benchmarking series that are highly volatile in sign and/or movements.