ABSTRACT
An important aspect of Benzécri’s French school of data analysis is the recoding of data prior to visualization by correspondence analysis (CA), a theme treated in detail in the book by Murtagh (2005). The method of CA is seen as a universal tool for visualizing data of any kind, once recoded into a suitable form. To understand what makes a data set suitable for CA, one has to consider the elements of a frequency table, which is the primary object on which CA is applicable, and which needs no pretransformation:
• Each cell of a frequency table is a count. • The row or column frequencies are expressed relative to their mar-
ginal totals as profile vectors. • The marginal frequencies of the table provide masses that weight
the row and column profiles in the measure of variance and in the dimension reduction.
