Title: Robust Correspondence Analysis: Theory and Data Analysis
 
Authors: Marco Riani, Anthony C. Atkinson, Aldo Corbellini, Francesca Torti

Abstract:
In this talk we introduce a robust form of correspondence analysis based on
the tools of robust statistics. This leads to the systematic deletion of
outlying rows of the table and to plots of greatly increased informativeness
The robust method requires that a specified proportion of the data be used in
fitting. To accommodate this requirement we provide an algorithm that uses a
subset of complete rows and one row partially, both sets of rows being chosen
robustly. We prove the convergence of this algorithm. We also discuss some
methods to adaptively select the optimal proportion of trimming which has to
be used. Using a variety of examples, coming from different sources, we show
how the application of the suggested method can easily identify subsets of
the sample which are behaving in a different way from the bulk of the data.
The robust correspondence analysis code has been written in MATLAB, is part
of the FSDA toolbox and can be downloaded from GitHub or MathWorks file
exchange.