ABSTRACT
Compositional data consist of sets of osbervations that add up to a constant, such as proportions that sum to 1, or percentages that sum to 100. Data such as these, called compositions, are prolific in chemistry, biochemistry and geology, where samples are analytically measured for their relative contents of different substances, for example oxides in chemistry, fatty acids in biochemistry and grain sizes in geology. But compositional data are also found in the social and economic sciences, such as daily time budgets that sum to 24 hours, family expenses on different items as proportions of total expenditure, or government budgets for different public services. Methods for visualizing compositional data are intimately related to correspondence analysis, since correspondence analysis also analyses the relative values in a data set. In this chapter we will look at methods for visualizing compositional data and show their relationship with correspondence analysis.
