Multivariate descriptive statistical analysis

Description

This is a well-written and interesting book about techniques for displaying multi- variate data. Although the examples are applications to socioeconomic research, it is claimed that the methods can also be applied to the social sciences, medicine, biology, and geography. The primary focus is on correspondence analysis, with other techniques such as canonical correlation, discriminant analysis, and cluster analysis discussed in this context. One could conclude from the absence of exercises that the book is not intended as a text, but it certainly could be used for a class if supplemented with problems. The main prerequisite is linear algebra, but some calculus is used, too, including matrix derivatives and Lagrange multipliers. The style is informal, with techniques presented often in terms of the analysis of a particular data set, and there are no theorems presented as such. There are, however, some mathematical derivation. This is a clear, carefully written discussion of correspondence analysis, a methodology which deserves to be more widely known.

Description

There is an excellent chapter which relates correspondence analysis to discriminant analysis and canonical correlation analysis. Another chapter discusses cluster analysis and includes an example in which clustering is combined with correspondence analysis. First the clustering partitions the data into homogeneous groups and then the corre- spondence plot shows how the groups differ on their response. For the reader with access to a program which manipulates matrices with facility and has good plotting facilities, it should be possible to implement the procedures fairly easily without recourse to the FORTRAN program.

Subjects

Series Statement

• Wiley series in probability and mathematical statistics.