Análisis de correspondencias simple : aplicación a un estudio de geografía electoral

Abstract

The analysis of correspondence developed by Benzecri comes within the frame of descriptive multidimentional statistics. Based on the principles of classic factorial analysis founded by Sperman and Thurstone, it has as its aim the extraction of the main axes from a large group of data which are difficult to perceive immediately. However the anlysis of correspondence converts to brute data into relative frequencies, that permits of the treatment in general of information of a qualitative nature at the same time quantitative expressed in units of different sizes. In a more precise way, the analysis of correspondence sums up the information contained in a table of contingencies referring to two groups of large sizes and taking into account the probabilistic character of the data to remedy their heterogenity. Affter transforming the elements of the table, the anlysis of correspondence gives a perfectly clear vision of two clusters of points when this is projected on to a sub-spaace of small dimensions (usually two) in such way as to maintain an important part of the information initially constituted. So, it deals with giving a coherent mea ning to the axes on which the points have been projected and taking advantage of the certain help in interpretation which the method provides.