Conclusions
Performing ATR measurements instead of transmission
measurements permits easier manipulation of yogurt
samples and reduces the effect of water on the IR
measurements.
Due to the heterogeneity of the sample population, it is
necessary to use hierarchical cluster analysis to select the
calibration data set, since cluster classification based on the
mid-IR spectra of yogurts is mainly related to the total
amount of carbohydrate, fat and protein, and therefore to
the energetic value of the yogurt in question.
Good estimations were obtained for all nutritional
parameters, including Ca, with the exception of fat content,
where the estimation technique may be impeded by the
presence of lactic acid bands in the IR region of where fatty
compounds absorb.
It should be noted that the results obtained by the
proposed method for most of the nutritional parameters
comply with the US FDA statuary values.
Upon comparing the results obtained using single and
extended models, which had different numbers of samples
in their calibration sets, it is clear that the inclusion of
additional samples affects the optimum spectral range to be
used when determining the nutritional parameters of the
yoghurts, and it also tends to increase the number of factors
required for an accurate analysis. This indicates that the
PLS–ATR–FTIR analysis of yogurt samples is not very
robust. Therefore, the technique requires further development,
such as the addition of more data, for example the
lactic acid content, or its application needs to be restricted
to samples produced in a similar way.