Time domain (or low resolution pulse) NMR can generate a range of rela
xation curves (CPMG, ER, PS, HSE, FID, etc.) which may vary depending
on the characteristic of the product being controlled-water content, h
ydration state, solid fat content, iodine number or even authenticity
of origin, Very often, the NMR signal is decomposed into a sum of expo
nential relaxation curves and the calculated NMR parameters (e.g. R-1,
R-2, M-0) are correlated with the property under study. Chemometric t
echniques, such as analysis of variance (ANOVA) and factor analysis, w
ere shown to be effective means of determining whether a given set of
NMR signals contains any interesting information before proceeding to
use fastidious and often uncertain signal decomposition procedures. Th
ese techniques were applied to NMR signals acquired for spreads and ge
latines with different compositions, to mixtures of a cation (Cu2+) an
d a ligand (tetraphenylporphin) and to glycine solutions at different
pH values. (C) 1997 John Wiley & Sons, Ltd.