APPLICATION OF MULTIVARIATE DISCRIMINANT ANALYSIS IN ENVIRONMENTAL POLLUTION PROBLEMS

Abstract

It is difficult to imagine any type of scientific inquiry in the modern world that does not involve the recording of observations on one or more types of objects. The objects may be things, people, natural or man-made events. The selected objects - air, water, white rats, model airplanes, biopsy, slides, patterns of response to complex stimulus situations, ability tests, brand selection behaviour, corporate financial activities - vary with the investigators discipline. Whatever their nature, the objects themselves are never measured in total. Rather, what is recorded are observations dealing with characteristics of the objects, such as suspended particulate matters, D.O., weight, wind velocity, cell diameters, speed or latency of response, number of correctly answered questions, specific brand chosen, previous year sales and so on. It is often the case that two or more characteristics (B.O.D., pH, D.O. etc.) will be measured at the same time on each object being studied. As a set of statistical techniques, multivariate discriminant data analysis is strategically neutral. Techniques can be used for many purposes in environmental engineering - ranging from the analysis of data obtained from rigidly controlled experiments to teasing out relationships assumed to be present in a large mass