Near-infrared (NIR) spectra are often pre-processed in order to remove syst
ematic noise such as base-line variation and multiplicative scatter effects
. This is done by differentiating the spectra to first or second derivative
s, by multiplicative signal correction (MSC), or by similar mathematical fi
ltering methods. This pre-processing may, however, also remove information
from the spectra regarding Y (the measured response variable in multivariat
e calibration applications). We here show how a variant of PLS can be used
to achieve a signal correction that is as close to orthogonal as possible t
o a given Y-vector or Y-matrix. Thus, one ensures that the signal correctio
n removes as little information as possible regarding Y. In the case when t
he number of X-variables (K) exceeds the number of observations (N), strict
orthogonality is obtained. The approach is called orthogonal signal correc
tion (OSC) and is here applied to four different data sets of multivariate
calibration. The results are compared with those of traditional signal corr
ection as well as with those of no pre-processing, and OSC is shown to give
substantial improvements. Prediction sets of new data, not used in the mod
el development, are used for the comparisons. (C) 1998 Elsevier Science B.V
. All rights reserved.