Highly Influential

3 Excerpts

- Published 1988 in Automatica

-The upper bound of the bias in the estimated parameters is derived in terms of the spectral norm. It is then established that the bias is related to the condition of the problem as well as the noise existing in the data. Conventional methods such as the generalized least squares (GLS), the instrumental variables (IV) and the extended matrix (EM) method seek to minimize bias by reducing the level of noise correlation in the data. The method presented in this paper approaches the bias reduction problem via the condition number, which acts as a multiply factor on the bias due to noise correlation. By improving the condition of the problem, it can be shown that bias in the estimated parameters is significantly reduced.

@article{Le1988BiasRI,
title={Bias reduction in parameter estimation},
author={Loc X. Le and W. J. Wilson},
journal={Automatica},
year={1988},
volume={24},
pages={825-828}
}