Maximum likelihood estimators for inverse problems with nuisance parameters
After a brief review of existing methods for dealing with nuisance parameters, maximum likelihood estimators (MLE) for the parameters of interest were developed which fully account for uncertainties in prior estimates of nuisance parameters and which are specifically tailored to inverse problems with complicated forward models. The noise and uncertainties in prior estimates are assumed to be normally distributed with known covariance, so the MLEs are generalized versions of least squares fitting. Expressions for uncertainties in estimates are provided. Two examples are given of diagnostic systems in nuclear fusion research for which the systematic accommodation of uncertainties in prior estimates of nuisance parameters can lead to significant improvements in routine data analysis and in assessment of diagnostic potential.
Bibliographic Reference: Report: JET-R(97)13 EN (1997) 29pp.
Availability: Available from the Publications Officer, JET Joint Undertaking, Abingdon, Oxon, OX14 3EA (GB)
Record Number: 199711719 / Last updated on: 1998-01-20
Original language: en
Available languages: en