Raman error statistics

Ultimately chemical images have the ability to show API adhering to other formulation excipients which is not possible by other sizing methods. Statistically, laser diffraction methods have less errors in the measurement as they measure millions of particle compared to the hundreds by Raman chemical imaging, but their lack of specificity means the whole sample is measured and not just the API. Ultimately the Raman method is still under development but provides the formulators of inhaled products an additional tool in understanding formulation properties and product. 

In the following sections, we review the application of Raman spectroscopy to glucose sensing in vitro. In vitro studies have been performed using human aqueous humor (HAH), filtered and unfiltered human blood serum, and human whole blood, with promising results. Results in measurement accuracy are reported in root mean squared error values, with RMSECV for cross-validated and RMSEP for predicted values. The reader is referred to Chapter 12 for discussion on these statistics. 

For all samples, except the vulcanisate of EPDM with 4.6 wt.% ENB and 1.5 phr sulfur, the ratio of the converted number of ENB moieties and the number of chemical crosslinks is around 2.0 ( 0.2). The statistical spread in the ratio of the converted number of ENB molecules and the number of chemical crosslinks is estimated to be 10-15% based on the error in the Raman and NMR data. Furthermore, there might be a systematic error in the crosslink density as determined by NMR, originating from the assumptions made in calculating the number of chemical crosslinks from the total number of crosslinks (chemical and entanglements). 

The fixed monitor count used in the data collection is chosen to optimise its statistical errors. Spectra are accumulated over several hours or overnight (or longer) for weak samples. On FANS a good signal to noise spectrum from co 8 g of a typical organic compound can be recorded in about five hours (excluding the time to cool the sample ( 3.5.2)). The errors in INS spectroscopy are, as in infi ared, Raman and NMR, governed by Poisson statistics, so the error on a data point of n counts, is fn This tyranny of the square root requires the measurement time to double before the statistics improve by only V2. ... 

---