Sample classification

The PLS calibration set was built mixing in an agate mortar different amounts of Mancozeb standard with kaolin, a coadjuvant usually formulated in agrochemicals. Cluster analysis was employed for sample classification and to select the adequate PLS model acording with the characteristics of the sample matrix and the presence of other components. 

The Dephy data-processing package offers several proprietary noise filtering options. We have not yet determined unequivocally whether any of these data treatments increase reproducibility or concentrate the taxonomic information content of the averaged spectra, although first results suggest that use of filtered data can improve spectral reproducibility and thus assist the ruggedness of sample classification. 

Developing a classification rule This step requires the known class membership values for all calibration samples. Classification rules vary widely, but they essentially contain two components ... 

Cluster analysis Is used to determine the particle types that occur in an aerosol. These types are used to classify the particles in samples collected from various locations and sampling periods. The results of the sample classifications, together with meteorological data and bulk analytical data from methods such as instrunental neutron activation analysis (INAA). are used to study emission patterns and to screen samples for further study. The classification results are used in factor analysis to characterize spatial and temporal structure and to aid in source attribution. The classification results are also used in mass balance comparisons between ASEM and bulk chemical analyses. Such comparisons allow the combined use of the detailed characterizations of the individual-particle analyses and the trace-element capability of bulk analytical methods. 

A sample classification is suspect if the K nearest neighbors are from multiple classes. 

Discriminant analysis (DA) performs samples classification with an a priori hypothesis. This hypothesis is based on a previously determined TCA or other CA protocols. DA is also called "discriminant function analysis" and its natural extension is called MDA (multiple discriminant analysis), which sometimes is named "discriminant factor analysis" or CD A (canonical discriminant analysis). Among these type of analyses, linear discriminant analysis (LDA) has been largely used to enforce differences among samples classes. Another classification method is known as QDA (quadratic discriminant analysis) (Frank and Friedman, 1989) an extension of LDA and RDA (regularized discriminant analysis), which works better with various class distribution and in the case of high-dimensional data, being a compromise between LDA and QDA (Friedman, 1989). 

Some common classification parameters are the mean values of the classes in the classification space, the variance of the class s calibration samples around the class mean, and the unmodeled variance in the calibration samples. Classification logic varies widely between classification methods. The following section provides details on some commonly encountered classification methods. 

Table 12.8 provides sample classifications for five "benchmark" chemicals, some of which are used in other chapters of this book. The table lists the substructure from Table 12.7 that determine the classification for the five chemicals. 

Stoyanova R, Nicholson JK, Lindon JC, Brown TR. Sample classification based on Bayesian spectral decomposition of metabonomic NMR data sets. Anal Chem 2004 76 3666-3674. 

Extensions to the earlier chemometric approaches include a toxicological assessment approach based on neural network software to ascertain whether the methods provide a robust approach, which could lead to automatic toxin classification. The neural network approach to sample classification, based on NMR spectra of urine, was in general predictive of the sample class. It appears to be reasonably robust and once the network is trained, the prediction of new samples is rapid and automatic. However, the principal disadvantage is common to all neural network studies in that it is difficult to ascertain from the network which of the original sample descriptors are responsible for the classification. Although recently it has been suggested that probabilistic... 

Li L, Weinberg CR, Darden TA, Pedersen LG. Gene selection for sample classification based on gene expression data study of sensitivity to choice of parameters of the GA/KNN method. Bioinformatics 2001 17 1131-42. 

Automated, rapid, high-throughput global analysis with minimal sample preparation, used for sample classification Limited ability for metabolite identification and quantification except for NMR... 

One of the prerequisites for automatic use of NMR spectra as descriptors to enable sample classification has been the need to reduce the spectra to a series of multidimensional coordinates. One way is to segment the spectrum and integrate over each segmented region, thereby removing the effects of minor chemical shift changes as a result of changes, e.g. in pH. ... 

Browse to the location of a map file (e.g., the map file supplied with Genome Viz for Escherichia coli K12 in the samples/classification-data directory - Escherichia coli K12.map). 

In this context, expert flow systems are important tools for screening purposes, as they permit real-time sample classification [370]. It should be emphasised that ordinary flow analysers used with the objective of offline classification of samples in different categories cannot be considered as expert flow systems. 

Sample classification use many GCxGC datasets to characterize sample classes based on within-class commonalities and between-class differences and then classify a sample into one of the classes based on GCxGC analysis. 

B., et d. (2004) Sample classification from protein mass spectrometry by peak probabflity contrasts . Bioinformatics, 20 (17), 3034-3044. 

The spectra are processed by computer software that now often includes advanced chemomet-ric methods (see the discussion of chemometrics in Section 4.7.2.1). Methods such as partial least squares (PLS) and principal component analysis are used by instrument software to construct calibration curves and provide quantitative analysis. LIBS software on many commercial systems supports library matching, sample classification, qualitative, semiquantitative and quantitative analysis including preloaded libraries and calibrations, and customer-built libraries and calibrations. 

Some further examples of sample classification from electronic tongues containing ICP-modified electrodes as at least one of the elements constituting the sensor array are reported in Table 2.4. 

On large data sets (greater than 14 samples) the SIMCA and Mahalanobis distance methods perform better than the wavelength distance method because they use estimates of the underlying population variance/covariance matrix for sample classification. As the size of the training set decreases, the accuracy of the population variance/covariance matrix estimates decreases, and performance of the Mahalanobis distance and SIMCA methods is reduced. In such cases where small training sets are used, the wavelength distance method may perform better because its univariate means and standard deviations are likely estimated with more certainty than the multivariate means and variance/covariance matrices used by the Mahalanobis distance and SIMCA methods. 

Fig. 4.4 A sample classification of selected supply chain configuration problems...

Including the sum squared spectral residual as an additional discriminating factor for the Mahalanobis group is an important extra step that improves the sensitivity of the unknown sample classifications. It not only sets the maximum allowed variation in the factors but also limits the range of variation in the residual for a sample to be classified as a member of the group. This is particularly important in quality control applications in which it is important not only to verify the identity of a material but also to determine if it contains substantial impurities different from those in the training data. 

An intrinsic decomposition reflects a well-founded relation selected via the Intrinsic Heuristic, and an extrinsic or logarithmic decomposition reflects a well-founded relation selected via the Extrinsic Heuristic (see Chapter 4). Sample classifications are given below. 

Other categories could be added here. Sample classifications are given below. 

---