Multiple correspondences and log-linear adjustment to test single correspondence relationships and categorization of qualitative data in e-commerce

When an exploratory data analysis is performed where there are more than two qualitative variables, the application of univariate, bivariate and multivariate statistical techniques allows the data table to be described successfully. Particularly, the single correspondence technique gives important correlation and dimensionality reduction results, which helps to give an objective interpretation of the data. We use the technique known as factor analysis of multiple correspondences, which is a generalization of the single correspondence technique used to corroborate results. We also use log-linear adjustment, with the purpose of continuing with the principal components and cluster analyses (Bernabe L.B and Olsina, L., Novena Conferencia de Ingenieria Electrica, 2003). The binary variables under study are the result of the e-commerce sites\´ evaluation process for the quality attributes of the "functionality" feature (Lafuente, G.H., et al., Proc. JUCSE 00, Nuevas Tendencias en Ingenieria de Software, 2000; Loranca, M.B. and Olsina, L., 6^o Workshop Iberoamericano de Ingenieria de Requisitos y Ambientes Software, p.178-89, 2003). This data is concentrated in a binary table of 49 sites and 17 attributes (Bernabe and Olsina, 2003; Lafuente et al., 2000).