Chemometrics review for chemical sensor development, task 7 report

Abstract

This report, the seventh in a series on the evaluation of several chemical sensors for use in the U.S. Department of Energy`s (DOE`s) site characterization and monitoring programs, concentrates on the potential use of chemometrics techniques in analysis of sensor data. Chemometrics is the chemical discipline that uses mathematical, statistical, and other methods that employ formal logic to: design or select optimal measurement procedures and experiments and provide maximum relevant chemical information by analyzing chemical data. The report emphasizes the latter aspect. In a formal sense, two distinct phases are in chemometrics applications to analytical chemistry problems: (1) the exploratory data analysis phase and (2) the calibration and prediction phase. For use in real-world problems, it is wise to add a third aspect - the independent validation and verification phase. In practical applications, such as the ERWM work, and in order of decreasing difficulties, the most difficult tasks in chemometrics are: establishing the necessary infrastructure (to manage sampling records, data handling, and data storage and related aspects), exploring data analysis, and solving calibration problems, especially for nonlinear models. Chemometrics techniques are different for what are called zeroth-, first-, and second-order systems, and the details depend on the form of the assumed functional relationship between the measured response and the concentrations of components in mixtures. In general, linear relationships can be handled relatively easily, but nonlinear relationships can be difficult.