Multiple correspondences and log-linear adjustment in e-commerce

Abstract

When an exploratory data analysis is performed where more than two qualitative variables are present, the application of univariate, bivariate and multivariate statistical techniques allows to successfully describe the dataset. Particularly, the single correspondence technique gives important correlation and dimensionality reduction results, which helps when giving an objective interpretation of the data. In this paper the technique known as factor analysis of multiple correspondences is used, which is a generalization of the single correspondence technique used to corroborate results. The log-linear adjustment is used too, with the purpose of continuing with the Principal Components and Cluster Analyses [3]. The binary variables under study are the result of the ecommerce sites' evaluation process for the quality attributes of the "Functionality" feature [8, 9]. These data is concentrated in a binary table of 49 sites and 17 attributes [3, 8]. (See table A.1 in the appendix for the list of variables). © Springer-Verlag 2004.