Search

A discussion of multivariate calibration techniques and their possible application to sensor array data is reviewed. The progression from multiple linear regression, and principal components regression, to partial least squares can be described by the analysis of their respective algorithms and the data analysis of their respective algorithms and the data analysis problems each of them solve. Additional analysis models are introduced to detect underlying background problems for sensor arrays and the ability to quantitate analysis directly by rank annihilation factor analysis when two dimensional data is obtained. Keywords include: Multivariate Analysis, Multisensor Arrays, Partial Least Squares (PLS), Multiple Linear Regression (MLR), Principal Component Regression (PCR), Rank Annihilation Factor Analysis (RAFA).

A systematic examination of calibration and quantitation methods available to analytical chemistry is made. First, simple linear calibration with one sensor is reviewed with an emphasis on chemical problems that can invalidate calibration models and what can be done about them. Then, shift is made to multivariate methods used for multicomponent analysis ending in a discussion of bilinear forms. (Author)

The multicomponent linear additive model is frequently used in analytical chemistry for the analysis of several components in a single sample. The pressure of unsuspected components in the mixture, which affect the measured response, results in a volume dependent sample background. A method has been developed to test for the presence of a background interferent during the quantitation step of a multicomponent analysis. If a background is detected, quantitative estimates of the desired analyte concentrations can be obtained by applying several chemically and physically meaningful constraints. Two distinct quantitation methods, which are known as the perpendicular projection and the extreme vertex projection techniques, have been proposed. (Author)

Theory of oscillations in backscattered ion intensity with incident ion energy in Ion Scattering Spectroscopy is reviewed and the analytical potential of this phenomena in surface analysis is demonstrated for a series of closely related lead compounds with the incident energy of the helium probe ion ranging from 500 eV to 4000 eV. These data show Oscillatory Ion Scattering Spectrometry to be very sensitive to the chemical environment of the target atom and so can be used to positively identify the chemical species where ISS cannot.

The detection and subsequent correction of drift in an analytical signal occurring during the course of an analysis is a common problem in analytical chemistry. In this paper, a solution to this problem is presented in the form of an extension to the theory of the Generalized Standard Addition Method (GSAM), a method currently used to simultaneously correct for matrix effects and interferences in multicomponent analysis. (Author)

A general-purpose interface for the PDP-11 family of minicomputers is described. The interface, oriented towards laboratory data-acquisition and experimental control applications, can be built in a relatively short time with low materials costs. A general purpose, FORTRAN-compatible software package capable of driving the interface is also discussed. (Author)

A procedure for including measured analytical uncertainty into data analysis methodology is discussed. The suitability of various dispersion matrices and matrix rank determination criteria for data having analytical uncertainty is investigated. A criterion useful for judging the number of factors insensitive to analytical uncertainty is presented. A model data structure for investigating the behavior of factor analysis techniques is described and analyzed. (Author)

Pattern recognition techniques are applied to the problem of determining the source of an oil spill. The utility of various elements for making this determination is assessed. Classification procedures using comparisons of the field sample to single known source samples and to multiple ('fuzzy') artificially weathered source samples of a given type are presented and discussed., Prepared in cooperation with Western Regional Research Lab., Berkeley, Calif.

Pattern recognition is used to predict the relative reactivity of a molecule from its structural features. For this preliminary study the hydrolysis of substituted benzoic acids was selected as the model reaction system. Nine structurel features of seventy-nine ortho, meta, or para substituted benzoic acids were input to a master program of pattern recognition methods. These methods were 96.2% accurate in predicting the relative reactivity of the substituted acids. The features that describe the electronic structure of the substituent and its position on the aromatic ring were found to be more important for predicting the rate than the features that describe the steric size of the substituent. (Author), See also AD-784 391.

A molecular structure transform has been developed that allows the three-dimensional structure of molecules to be related to their biological activity. The transforms utilize principal component analysis to transform the three dimensional cartesian coordinates and any atomic information such as electronegativity for each atom in a molecule. A predetermined and constant number of features is extracted from each molecule which can be analyzed by the methods of pattern recognition in order to relate molecular structure to any molecular function; in this case, carcinogenicity. The transform is graphically demonstrated using a simple triatomic molecule and then applied to two divergent classes of carcinogenic molecules; dimethylaminoazobenzenes and polycyclic aromatic hydrocarbons. Structure-activity results are presented along with an exciting potential for drug design. (Author)

A new preprocessing method that has found considerable use for chemical applications is presented. Features are selected that are useful for class discrimination, orthogonal, and at least in part, retain identity to the original measurements. A brief introduction to chemical pattern recognition is presented followed by a description of the method and an application to mass spectral data analysis. (Author)