ABSTRACT
Apart from the passing mention of the chi-square test, the discussion about significant clustering in Chapter 15 and the visualization of confidence ellipses in Chapter 28, this book has concentrated exclusively on the geometric properties of correspondence analysis (CA) and its interpretation. In these two final chapters we explain some approaches to statistical inference in the context of CA and related methods. In the present chapter we shall be investigating the stability of CA solutions and the sampling properties of statistics such as the total inertia, principal inertias and principal coordinates. The distinction is made between (i) stability of the solution, irrespective of the source of the data, (ii) sampling variability, assuming the data arise out of some form of random sampling from a wider population, and (iii) testing specific statistical hypotheses.
