Experiencias con un nuevo índice de falta de ajuste en Tablas de Contingencia

Abstract

Rudas, Clogg and Lindsay, en 1994, proposed a new index of lack of fit for contingency table analysis. For a contingency table P with N cells, and a model H, the new index is defined as the smallest \pi that satisfies the equation P_h=(1-\pi)\Pi_{1h}+\pi\Pi_{2h}, h=1,...,N with 0 \leq \pi \leq 1, and \Pi_1 and \Pi_2 the tables of probabilities for each latent class. The model H applies to \Pi_1 but not impose any restrictions on \Pi_2. The interpretation of the index of lack of fit is the proportion of individuals of the population intrinsically outside model H. In this work we propose the analysis of the second latent class to the study of the causes of lack of fit of a model and to detect outlying cells. At the same time, we propose the new index of lack of fit as a concepto of non-independence in the two simple configural frequency analysis