A generalized partial-least-squares calibration for determination of the tr
ans content of edible fats and oils by Fourier transform near-infrared (FT-
NIR) spectroscopy using 8-mm disposable glass vials for sample handling and
measurement was developed. The trans contents of a broad range of oils wer
e determined using the American Oil Chemists' Society single-bounce horizon
tal attenuated total reflectance (SB-HATR) mid-infrared spectroscopic proce
dure, these trans reference data were used in the development of the genera
lized FT-NIR calibration. Additional refined and product-specific calibrati
ons were also developed, and all the calibrations were assessed for their p
redictive capabilities using two sets of validation samples, one comprising
a broad range of oil types and the other restricted to oils with specific
characteristics. The FT-NIR trans predictions obtained using the generalize
d calibration were in good agreement with the SB-HATR results; the values w
ere accurate and reproducible to within +/-1.1 and +/-0.5% trans, respectiv
ely, compared to a reproducibility of +/-0.40% trans obtained for the SB-HA
TR method. The accuracy of the predictions obtained from the generalized FT
-NIR calibration for particular oil types was not significantly improved by
supplementing the base training set with samples of these specific types.
Calibrating only these oil types did, however, produce a substantial improv
ement in predictive accuracy, approaching that of the SB-HATR method. These
product specific calibrations produced serious predictive errors when nonr
epresentative samples were analyzed. The incorporation of a supplementary d
iscriminate analysis routine was found to be a powerful safeguard in flaggi
ng nonrepresentative samples as outliers and could also be used to select t
he calibration most appropriate for the characteristics of the sample being
analyzed. Overall, it was concluded that FT-NIR spectroscopy provides a vi
able alternative to the SB-HATR/mid-Fourier transform infrared method for t
rans determination, making use of more industrially robust instrumentation
and equipped with a simpler sample handling system.