On Correspondence Analysis of Incomplete Orderings

Correspondence analysis (CA) is a method for graphical presentation of two-way data tables. We consider the following problem: how can CA be applied in a particular case when the data consists of incomplete orderings. As an example of this kind of data we refer to a contest where each member of the jury allocates ranks 1,2, …, k only to k “best” candidates (out of n). In order to apply CA for such data a restricted iterative proportional fitting procedure is proposed to modify the initial table of ranks by imputing nonzero numbers to cells that were not among the k best cells. We demonstrate the work of the method via a practical example.