IDENTIFYING PROCESS MODELS WHEN SOME DATA ARE MISSING - 2 OPTIONS AREHIGHLIGHTED

Authors
ISAKSSON AJ KAUL V
Citation
Aj. Isaksson et V. Kaul, IDENTIFYING PROCESS MODELS WHEN SOME DATA ARE MISSING - 2 OPTIONS AREHIGHLIGHTED, Pulp & paper Canada, 96(8), 1995, pp. 48-51
Citations number
9
Categorie Soggetti
Materials Science, Paper & Wood
Journal title
Pulp & paper Canada → ACNP
ISSN journal
03164004
Volume
96
Issue
8
Year of publication
1995
Pages
48 - 51
Database
ISI
SICI code
0316-4004(1995)96:8<48:IPMWSD>2.0.ZU;2-C
Abstract
This paper reviews and compares some identification methods applicable to situations where some of the data are missing. Based an Monte Carl o simulations, but also on real data from a pulp mill, the authors sho w that linear interpolation can lead to severly biased estimates. The only two viable options are the EM algorithm and the least-squares met hod ignoring time instants with incomplete information. The end choice is a trade-off between accuracy and computational complexity.