Because of the low active component concentration in the studied pharmaceuticals, the intensities of the measured spectra were weak. It means that the spectra obtained were characterised by a rather low signal-to-noise (S/N) value of the order of 30. In the case of diclofenac sodium solutions the principal component analysis (PCA) showed that in such a complex system only about 85% of the spectral variation could be accounted for by the first five principal components. The next principal components connected with the noise and fluctuations of the baseline decreased slowly. After smoothing of the spectra, the first five principal components accounted for nearly 94% of the spectral variation in the studied system, however smoothing had negligible influence on the parameters of the elaborated calibration models.At the beginning, isolated diclofenac peaks at 1581 and 1604 cm−1 were chosen to perform analysis in a classical way, using a univariate approach. It was necessary to assume that the remaining substances, present in the studied systems, did not interfere strongly in this spectral region. Band intensities and band areas were calculated applying one-point baseline correction. As one could expect, univariate calibration models were of low quality. The determination coefficient, R2, for obtained calibration curves equalled 0.852 or 0.796 (Table 1). The relative standard error of prediction determined for the testing data set amounted to 11.5 and 13.7%, for the model based on band area and band intensity, respectively.
Because of the low active component concentration in the studied pharmaceuticals, the intensities of the measured spectra were weak. It means that the spectra obtained were characterised by a rather low signal-to-noise (S/N) value of the order of 30. In the case of diclofenac sodium solutions the principal component analysis (PCA) showed that in such a complex system only about 85% of the spectral variation could be accounted for by the first five principal components. The next principal components connected with the noise and fluctuations of the baseline decreased slowly. After smoothing of the spectra, the first five principal components accounted for nearly 94% of the spectral variation in the studied system, however smoothing had negligible influence on the parameters of the elaborated calibration models.<br><br>At the beginning, isolated diclofenac peaks at 1581 and 1604 cm−1 were chosen to perform analysis in a classical way, using a univariate approach. It was necessary to assume that the remaining substances, present in the studied systems, did not interfere strongly in this spectral region. Band intensities and band areas were calculated applying one-point baseline correction. As one could expect, univariate calibration models were of low quality. The determination coefficient, R2, for obtained calibration curves equalled 0.852 or 0.796 (Table 1). The relative standard error of prediction determined for the testing data set amounted to 11.5 and 13.7%, for the model based on band area and band intensity, respectively.
正在翻译中..