Kinds Of Multivariate Methods

There are at least 14 kinds of MVA procedures used to examine sensory-instrumental data. Some are of chief benefit for exploratory examination of data to gleam from large data sets certain components or relations for more detailed examination or later use. Others are employed to learn whether sample differences exist. Among the procedures are

Principal component analysis Cluster analysis Multidimensional scaling Discriminant analysis Multiple regression analysis Partial least squares analysis Response surface analysis Procrustes analysis Canonical analysis Correspondence analysis Multivariate analysis of variance Fuzzy logic (mathematics) Neural network Factor analysis

Of the above procedures, principal component analysis (PCA), cluster analysis (CA), and factor analysis (FA) are known as methods of internal analysis. They are used to study the relations of variables within the same data set, ie, a set of sensory measurements or a set of instrumental measurements, but not sensory compared with instrumental measurements. Discriminant analysis (DA) and multiple regression (MRA) perform just the opposite functions. DA is a procedure for assigning products, or some other entity, to a class based on measurements made on those products, with prior knowledge that classes do exist. For example, yellow and red globe onions and Vidalia (Grano-Granex) onions were analyzed gas chromatographically to learn if they could be so classified (7). That species differences existed was, of course, known in advance. MRA works somewhat analogously to DA except it applies to things that progress in some manner rather than things in discrete classes, as species of onions are. An illustration would be predicting the intensity of vanilla flavor in ice cream when different levels of vanilla are added. DA and MRA are the most useful of the MVA methods for predicting the identity or state of a sample. They are useful because once a discriminant function or a multiple regression equation has been calculated, the identity or state of a new sample can be predicted by merely inserting its measurement values in the equation. The other methods listed above are more commonly used to study relations between sensory and instrumental measurements. Some, such as canonical analysis or partial least squares regression can be used for prediction purposes, but generally they are used to examine relations among variables rather than to pinpoint the identity of a single sample.

600 Chocolate Recipes

600 Chocolate Recipes

Within this in cookbook full of chocolate recipes you will find over 600 Chocolate Recipes For Chocolate Lovers.

Get My Free Ebook

Post a comment