Variables chemometrics

V-M Taavitsainen and P Korhonen. Nonlinear data analysis with latent variables. Chemometrics Intell. Lab. Sys., 14 185-194, 1992. 

To this point, the discussion of regression analysis and its applications has been limited to modelling the association between a dependent variable and a single independent variable. Chemometrics is more often concerned with multivariate measures. Thus it is necessary to extend our account of regression to include cases in which several or many independent variables contribute to the measured response. It is important to realize at the outset that the term independent variables as used here does not imply statistical independence, as the response variables may be highly correlated. 

Partial least-squares path modeling with latent variables (PLS), a newer, general method of handling regression problems, is finding wide apphcation in chemometrics. This method allows the relations between many blocks of data ie, data matrices, to be characterized (32—36). Linear and multiple regression techniques can be considered special cases of the PLS method. 

Other chemometrics methods to improve caUbration have been advanced. The method of partial least squares has been usehil in multicomponent cahbration (48—51). In this approach the concentrations are related to latent variables in the block of observed instmment responses. Thus PLS regression can solve the colinearity problem and provide all of the advantages discussed earlier. Principal components analysis coupled with multiple regression, often called Principal Component Regression (PCR), is another cahbration approach that has been compared and contrasted to PLS (52—54). Cahbration problems can also be approached using the Kalman filter as discussed (43). 

In the context of chemometrics, optimization refers to the use of estimated parameters to control and optimize the outcome of experiments. Given a model that relates input variables to the output of a system, it is possible to find the set of inputs that optimizes the output. The system to be optimized may pertain to any type of analytical process, such as increasing resolution in hplc separations, increasing sensitivity in atomic emission spectrometry by controlling fuel and oxidant flow rates (14), or even in industrial processes, to optimize yield of a reaction as a function of input variables, temperature, pressure, and reactant concentration. The outputs ate the dependent variables, usually quantities such as instmment response, yield of a reaction, and resolution, and the input, or independent, variables are typically quantities like instmment settings, reaction conditions, or experimental media. 

Theoretically based correlations (or semitheoretical extensions of them), rooted in thermodynamics or other fundamentals are ordinarily preferred. However, rigorous theoretical understanding of real systems is far from complete, and purely empirical correlations typically have strict limits on apphcabihty. Many correlations result from curve-fitting the desired parameter to an appropriate independent variable. Some fitting exercises are rooted in theory, eg, Antoine s equation for vapor pressure others can be described as being semitheoretical. These distinctions usually do not refer to adherence to the observations of natural systems, but rather to the agreement in form to mathematical models of idealized systems. The advent of readily available computers has revolutionized the development and use of correlation techniques (see Chemometrics Computer technology Dimensional analysis). 

The variable selection methods have been also adopted for region selection in the area of 3D QSAR. For example, GOLPE [31] was developed with chemometric principles and q2-GRS [32] was developed based on independent CoMFA analyses of small areas of near-molecular space to address the issue of optimal region selection in CoMFA analysis. Both of these methods have been shown to improve the QSAR models compared to original CoMFA technique. 

Principal Component Analysis (PCA). Principal component analysis is an extremely important method within the area of chemometrics. By this type of mathematical treatment one finds the main variation in a multidimensional data set by creating new linear combinations of the raw data (e.g. spectral variables) [4]. The method is superior when dealing with highly collinear variables as is the case in most spectroscopic techniques two neighbor wavelengths show almost the same variation. 

A more recently introduced technique, at least in the field of chemometrics, is the use of neural networks. The methodology will be described in detail in Chapter 44. In this chapter, we will only give a short and very introductory description to be able to contrast the technique with the others described earlier. A typical artificial neuron is shown in Fig. 33.19. The isolated neuron of this figure performs a two-stage process to transform a set of inputs in a response or output. In a pattern recognition context, these inputs would be the values for the variables (in this example, limited to only 2, X and x- and the response would be a class variable, for instance y = 1 for class K and y = 0 for class L. 

Wold, S., Kettaneh, N., and Tjessem, K., Hierarchical multiblock PLS and PC models for easier model interpretation and as an alternative to variable selection, J. Chemometrics 10, 463 (19%). 

In addition to the measured values and the analytical values (e.g. content, concentration), latent variables are included in the scheme. Latent variables can be obtained from measured values or from analytical values by means of mathematical operations (e.g. addition, subtraction, eigenanal-ysis). By means of latent variables and their typical pattern (represented in chemometric displays) special information can be obtained, e.g. on quality, genuineness, authenticity, homogeneity, origin of products, and health of patients. 

All those discussions, however, were based on considerations of the effects of multiple sources of variability, but on only a single variable. In order to compare Statistics with Chemometrics, we need to enter the multivariate domain, and so we ask the question Can ANOVA be calculated on multivariate data The answer to this question, as our long-time readers will undoubtedly guess, is of course, otherwise we wouldn t have brought it up ... 

The second critical fact that comes from equation 70-20 can be seen when you look at the Chemometric cross-product matrices used for calibrations (least-squares regression, for example, as we discussed in [1]). What is this cross-product matrix that is often so blithely written in matrix notation as ATA as we saw in our previous chapter Let us write one out (for a two-variable case like the one we are considering) and see ... 

Chapters 3 6 deal with direct mass spectrometric analysis highlighting the suitability of the various techniques in identifying organic materials using only a few micrograms of samples. Due to the intrinsic variability of artefacts produced in different places with more or less specific raw materials and technologies, complex spectra are acquired. Examples of chemometric methods such as principal components analysis (PCA) are thus discussed to extract spectral information for identifying materials. 

A total of 185 emission lines for both major and trace elements were attributed from each LIBS broadband spectrum. Then background-corrected, summed, and normalized intensities were calculated for 18 selected emission lines and 153 emission line ratios were generated. Finally, the summed intensities and ratios were used as input variables to multivariate statistical chemometric models. A total of 3100 spectra were used to generate Partial Least Squares Discriminant Analysis (PLS-DA) models and test sets. 

This book is the result of a cooperation between a chemometrician and a statistician. Usually, both sides have quite a different approach to describing statistical methods and applications—the former having a more practical approach and the latter being more formally oriented. The compromise as reflected in this book is hopefully useful for chemometricians, but it may also be useful for scientists and practitioners working in other disciplines—even for statisticians. The principles of multivariate statistical methods are valid, independent of the subject where the data come from. Of course, the focus here is on methods typically used in chemometrics, including techniques that can deal with a large number of variables. Since this book is an introduction, it was necessary to make a selection of the methods and applications that are used nowadays in chemometrics. 

Highly correlating (collinear) variables make the covariance matrix singular, and consequently the inverse cannot be calculated. This has important consequences on the applicability of several methods. Data from chemistry often contain collinear variables, for instance the concentrations of similar elements, or IR absorbances at neighboring wavelengths. Therefore, chemometrics prefers methods that do not need the inverse of the covariance matrix, as for instance PCA, and PLS regression. The covariance matrix becomes singular if... 

An essential concept in multivariate data analysis is the mathematical combination of several variables into a new variable that has a certain desired property (Figure 2.14). In chemometrics such a new variable is often called a latent variable, other names are component or factor. A latent variable can be defined as a formal combination (a mathematical function or a more general algorithm) of the variables a latent variable summarizes the variables in an appropriate way to obtain acertain property. The value of a latent variable is called score. Most often linear latent variables are used given by... 

Also nonlinear methods can be applied to represent the high-dimensional variable space in a smaller dimensional space (eventually in a two-dimensional plane) in general such data transformation is called a mapping. Widely used in chemometrics are Kohonen maps (Section 3.8.3) as well as latent variables based on artificial neural networks (Section 4.8.3.4). These methods may be necessary if linear methods fail, however, are more delicate to use properly and are less strictly defined than linear methods. 

Multivariate calibration has the aim to develop mathematical models (latent variables) for an optimal prediction of a property y from the variables xi,..., jcm. Most used method in chemometrics is partial least squares regression, PLS (Section 4.7). An important application is for instance the development of quantitative structure—property/activity relationships (QSPR/QSAR). 

In chemometrics, the letter p is widely used for loadings in PCA (and partial least-squares [PLS]). It is common in chemometrics to normalize the lengths of loading vectors to 1 that means p p = 1 m is the number of variables. The corresponding... 

---