Reference analysis

Figure. 3.20 Colocation of correct upstream sampler and PAT sensor, shown here for a recirculation loop configuration (right). The PAT sensor is deployed in upstream small-diameter piping with a field of view matching as best possible the same stream segment which is being sampled for reference analysis. This setup allows for reliable multivariate calibration because of the colocated [X, f] modalities.

The test set method of validation is rather straightforward, but requires some caution. First, the test set samples must be sufficiently representative of samples that the model will be applied to in the future. Otherwise, external validation can provide misleading results - in either an optimistic or pessimistic manner. Under-representing the anticipated sample states could lead to optimistic results, while the use of sample states not represented in the calibration data could lead to pessimistic results. A practical disadvantage of the test set method is that it requires that the reference analytical method be performed on an additional set of samples beyond the original calibration samples, which can be an issue if the cost of each reference analysis is high. 

Improved reference analysis If the method assessment reveals that the main source of model error comes from error in the reference analytical method used to generate the (y) calibration data, then efforts to improve the accuracy and precision of this method could prove to be very beneficial. 

The spinning solutions are not stable over time, since CS2 continues to react with the cellulose, increasing the xanthogenate content or ripeness (which is why monitoring is needed in the first place). This was dealt with by carrying out the NIR and reference methods simultaneously, where possible, or by storing the samples at -30 °C (where they are stable) until the reference analysis could be performed. 

Near Infrared Reflectance Analysis (NIRA) is in use at over 5000 sites for the analysis of multiple constituents in food and other products. The technology is based upon correlation transform spectroscopy, which combines NIR spectrophotometry and computerized analysis of a "learning set" of samples to obtain calibrations without the need for detailed spectroscopic knowledge of factors being analyzed. The computer can obtain spectral characteristics of the analyte (based upon a correlation with data from an accepted reference analysis) without separation of the sample s constituents. 

While there will be much more extensive discussions in later chapters on the topic of analyzer calibration, I will try to give you a quick snapshot here on how NIR analysis is done. The first step is to have a valid sample set, normally 30-50 samples covering the range of interest. The spectra are collected and a correlation is built between the wavelengths and their corresponding absorbencies and the reference analysis of these samples. The chemometric tools used in developing calibrations are discussed in Chapter 8. This is just the first step in developing a calibration but many people make the mistake of... 

They showed that the dimerized bond should be considered as a singlet diradical. Therefore, a qualitatively correct description of the dimer requires at least a generalized valence bond, GVB-PP132, or a two configuration self-consistent field (TCSCF)33 wave function. More recently, Paulus29 performed a more exhaustive multi-reference analysis of silicon clusters and reconfirmed this conclusion. 

The reference analysis of the electroiuc structure of molecules and crystals, as coming from experiment or calculation, is due to Richard Bader [11], who showed how chemical concepts related to Atoms can be recognized from a pure physical observable, like the electronic density, its derivatives, and its critical points. In the present study, Bader s analysis has been applied to porphyrin and metal porphyrins, mainly for calculating atomic charges. Moreover, we have explored how the electronic density analysis describes the metal-ligand bond (Figs. 1, 2 and 3 Table 1). 

Why do the mixtures exhibit a small heat generation The amount of Fe metal in the electrode was estimated to be c. 0.3 mg based on 1 mg of dischiu ged FeFs electrode in the conversion state. Moreover, the Fe metal generated in the discharged electrode after the conversion reaction has a rather small particle size. Therefore, 0.3 mg of nanosized Fe powder was used for the reference analysis. Figure 20.17 shows the DSC curves of... 

The reference analysis methods of EPA are given as numbers— 502.2 and 524.2. The method 502.2 is analysis of VOCs in water using capillary gas chromatography-photoionization detector-electron capture detector (CGC-PID-ELCD) and 524.2 is the method for VOCs using CGC-MS [44]. 

A short note on reference analysis The very first step in application of NIRS to analysis of a commodity should be that of determination of the error of the reference methods used in development of calibration models. This is the starting point of any NIRS analysis. Calibration model development and ongoing monitoring of accuracy and precision of analysis will always have to rely on reference analysis for verification of results. One of the serendipitous results of NIRS application worldwide has been the discovery in many operations of the need for improvement in their own hitherto sacrosanct wet chemistry laboratory practices. 

Table 7.3.2 gives an overview of the chemical reference analysis. Reflectance NIR measurements gave better prediction results compared to transmittance NIR measurements. Transmittance measurements will not be discussed further. The lowest prediction error for reflectance measurements for pork was 0.60% for moisture, 0.66% for protein, and 0.16% for fat (Table 7.3.3). The latter is probably somewhat overoptimistic, because the standard error of calibration was significantly higher, that is, SEC = 0.27%. The corresponding prediction errors for beef were comparable, namely, 0.52%, 0.61%, and 0.31% for moisture, protein, and fat, respectively. By...

Table 7.3.4 gives an overview of the chemical reference analysis. 

For most applications, it is essential that the state of all samples be identical when the results of the reference analysis are used in calibration. This means that physical parameters such as particle size and shape of the calibration test, and reference samples should be mutually compatible, and that the reference results must be relayed to the NIR instrument on the same as-is moisture, oil, or more simply constitutional basis that the instrument will see. The formula for conversion of the percentage of a constituent to an as-is moisture basis from data reported on a constant moisture basis is ... 

In the case of NIR instruments, sample selection is really a factor only in calibrations and their validation, because the operator has no choice in the samples presented for routine analysis. Calibration has been a bone of contention ever since the dawn of the NIR era. The most serious criticism of NIR technology is that the instruments require separate calibrations for different commodities for the same constituent, as well as for different constituents of the same commodity. There are two main methods of calibration. The first calls for accumulation of sufficient samples with reference analyses to enable generation of calibration and verification files, followed by selection of samples to fill these files. The second method involves selection of samples strictly on the basis of spectral characteristics, followed by reference analysis on only the relatively small number of samples that display the most comprehensive variance in optical data. These two methods are referred to respectively as conventional and spectral methods of sample selection in this chapter. 

Modern NIR software, such as WINISI offers excellent methods for spectral sample selection, and is another procedure, which is likely to become much more widely used in the future. The philosophy is that when two samples have spectra that are very closely similar there is only a need for one of them. In this way, Gaussian distribution of calibration samples and the Dunne effect are largely eliminated, because many samples will be spectrally similar, and only a relatively small proportion of a population will be selected. This will also reduce the need for expensive reference analysis. 

The theory and principles of NIR are not just the physics and electronics. The theory and principles must include chemometrics and the reference analysis as part of the total technique. Lack of structural interpretative value has been partly compensated by chemometric evaluation techniques [225-228]. Chemometrics (after 1985) has enabled use of diffuse reflectance techniques on granular solids that produce spectra that could not be treated with a simple one-wavelength Beer s law approach. 

LIBS is a suitable method for on-line analysis of the elemental composition of recycled complex ther-moplasts from consumer electronics (ABS, PA, PC, PS, SBR, PPG, TPO, PVC, PPO/PS). The process analysis of such thermoplasts by LIBS (Nd YAG at 266 nm) was reported [139] data analysis consisted of multivariate methods and variable subset selection via a genetic algorithm. Reference analysis was... 

Rock Age Locality Pb/ Pb 207pb/204pb 208pb/204pb Type of Reference analysis... 

Tables 21 and 22 introduce the literature from heavy metals in and toxic materials in loeal soils reported by The Institute for Cleaning, Tokyo, in 1978 and 1980. From these references, analysis of eontents of the materials used for disposable diapers was deemed neeessary. Thus, microanalysis of heavy metals that are contained in the constituent materials of disposable diapers was requested by the Japan Environmental Health Center, Inc., and the results are shown in Table 23 (February 1986).

Keyword— Reference Analysis, Recurrent Breast cancer, Poisson process, Bayesian Inference. 

---