A. Brink et T. Westerlund, THE JOINT PROBLEM OF MODEL STRUCTURE DETERMINATION AND PARAMETER-ESTIMATION IN QUANTITATIVE IR SPECTROSCOPY, Chemometrics and intelligent laboratory systems, 29(1), 1995, pp. 29-36
A method for automatically selecting the wave numbers best suited for
quantitative analysis as well as for simultaneously estimating the par
ameters in the emerging model is presented. As an indicator of the goo
dness of the model, Akaike's information theoretic criterion (AIC) is
used. Since this approach involves the maximum likelihood estimate of
the parameters, the problem of how to scale the data prior to the calc
ulations is eliminated. The method described in this paper is not rest
ricted to Fourier transform infrared (FTIR) problems, but can be appli
ed to other similar problems, where both the model structure and the p
arameters should be determined. During the calibration stage, a mixed-
integer nonlinear programming problem must be solved. It is demonstrat
ed that the use of such modern optimization techniques makes it possib
le to solve these types of problems without tremendous computational e
ffort. During the prediction stage the obtained model is easy to use.