8. Patterns suggested by data
8.1 Samples as mixtures149
8.2 Determining the minimum number of factors151
8.3 Application to the Atlantic Rocks dataset155
8.4 Spiky factors156
8.5 Time-Variable functions160
Patterns Suggested by Data explores linear models that characterize data as a mixture of a few significant patterns, whose properties are determined by the data themselves (as contrasted to being imposed by the analyst). The advantage to this approach is that the patterns are drawn from the data themselves and may bring out features that reflect the physical processes of the system. The methodology, which goes by the names factor analysis and empirical orthogonal function (EOF) analysis is applied to a wide range of data ...