In the first part of the paper, the optimal estimator for normally nonmeasu
red primary outputs from a linear, time invariant and stable dynamic system
is developed. The optimal estimator is based on all available information
in known inputs and measured secondary outputs. Assuming sufficient experim
ental data, the optimal estimator can be identified by standard system iden
tification (SI) methods, utilizing an output error (OE) model. It is then s
hown that least squares estimation (LSE) and multivariate calibration by me
ans of principal component regression (PCR) or partial least squares regres
sion (PLSR) can be seen as special static cases of such a dynamic SI. Final
ly, it is shown that dynamic system PCR and PLSR solutions can be developed
as special cases of the general estimator for dynamic systems. (C) 1998 El
sevier Science B.V. All rights reserved.