ABSTRACT
Discriminant analysis or pattern recognition techniques can be used to identify the similarities and differences in different types of samples providing sample-sample comparisons. Discriminant analysis techniques are categorized according to the underlying assumptions used in the development of the models. Discriminant analysis techniques can be applied to both chemical and sensory data. For both types of data, the goal of a discriminant analysis is to be able to place a given sample in the group to which it is most similar. Chemical data are not well behaved in the statistical sense. This situation is a result of the measurement techniques used for sample characterization, biochemical, process, or formulation induced variations, and the inability to properly control all experimental variables during data collection. Model development is inherently related to feature selection in that the selected variables are tested for their discrimination ability in order to identify the combinations that yield the best results.
