factor analysis of a set of contingency tables

Abstract

[EN] The aim of this paper is to develope a factor method for the simultaneous analysis of several contingency tables. As in multiple correspondence analysis, the method studies similarity among rows, among columns and also the relationships between rows and columns. In the simultaneos analysis two rows are said to be similar if they are similar in the correspondence analysis of each table. With the poposed method it is possible to analyze several real contingency tables, and those belonging to the same survey in different countries or different years, and different surveys over the same population. The method tries to find the general trends in the tables, without allowing one of them predominate over the others. The method extends to the analysis of a set of surveys if the data over the different groups of people are coded in complete disjunctive tables juxtaposed over the modalities and is also extendable to factor analysis of incomplete disjunctive tables with conditional questions.

[ES] El objetivo de este trabajo consiste en desarrollar un método factorial para el análisis conjunto de varias tablas de contingencia que permita de forma semejante al análisis de correspondencias, el estudio de la semejanza entre el conjunto de las filas, de las columnas y las relaciones entre ambos conjuntos. En este análisis conjunto se considera que dos filas son semejantes si lo son en los análisis de las tablas separadas.