Infrared error statistics

CVt = 0.25/1.96 = 0.128. The number 0.128 is the largest true precision for a net error at +25% at the 95% confidence level. The number 1.96 is the appropriate t - statistic from the t distribution at the same confidence level. Since the coefficient of variation of pump error is assumed to be 5%, a method should have a CV analysis <0.102 to meet the CV accuracy standard. Tables IV and V7 shows that the infrared technique meets this requirement. 

Most statistical tests assume diat die underlying samples and experimental errors fall on a normal distribution. In some cases diis is not so for example, when analysing some analytical signals it is unlikely diat die noise distribution will be normal it is often determined by electronics and sometimes even data preprocessing such as die common logarithmic transform used in electronic absorption and infrared spectroscopy. 

Figure 11 Average values (n = 6, error bars = SD) of the shift in the peak maximum of the C-H asymmetric stretching absorbance plotted as a function of time during the ethanol liquid treatment protocol. ANOVA followed by Scheffe s F-test revealed a statistically significant difference P < 0.05) between the location of the peak maximum before and after the 30-min ethanol exposure. There was no statistically significant difference between the baseline value and that at 24 hr. (From Ref. 125. Reprinted from Journal of Controlled Release, 16, Bommannan et al. Examination of the effect of ethanol on human stratum comeum in vivo using infrared spectroscopy, pp. 299-304, 1991, with kind permission from Elsevier Science, NL, Sara Burgerhartstraat 25, 1055 KV, Amsterdam, The Netherlands.)...

Near infrared has been used to quantify amorphous content in a crystalline matrix [64], mixtures of multiple polymorphic forms [65], and physical mixtures of crystalline and amorphous drug in the presence of excipients [66,67]. Otsuka et al. [68] found that Fourier transform (FT)-NIR outperformed x-ray powder diffraction (XRPD) with respect to accuracy statistics over the calibration range and down to 1% (w/w). Other quantitative studies have reported prediction errors of 5 to 6% (w/w) of y-IMC (indomethacin) in a sample matrix including amorphous and a-IMC [69]. These results demonstrate NIR selectivity of the solid-state forms of IMC. Selectivity, the proportion of analyte signal unaffected by other spectral interferences, is especially critical for monitoring dispersions where physical instability could manifest in a combination of amorphous and crystalline forms (including various polymorphs). Similar studies were performed for lactose and showed the possibility to quantify down to 0.5% of crystalline material in an amorphous base [70]. 

A recombinant strain of E. coli was used in the development of a quantitative spectroscopic method for the measurement of poly(3HB) content in the cell (Kansiz et al. 2000). Fourier transform infrared (FTIR) spectroscopy and multivariative statistical methods were performed to determine the cellular poly (3HB) content, and a correlation coefficient of 0.988, with a standard error of 1.49 % poly(3HB), was obtained between the measured and the predicted values. The spectroscopic method circumvented the need to use solvents to extract the PHA before quantitative measurements. 

---