Modelo de Ecuación Estructural con Intervención de Variables Ordinales. Cálculo de sus Correlaciones

Abstract

In this paper, we present a description of the multivariate technique called Structural Equation Modeling (SEM) and an overview of the calculation of the correlation coefficients using ordinal variables. The Structural Equation Modeling technique allows us to explain the relations between a set of variables observed in terms of a smaller number of unobserved variables, called factors or latent variables, which generate the structure or pattern of the relations among the former. This technique has been applied to social studies and market studies Price [1], Chau [2]; but the application on industrial process is just beginning. López and Carrión [3] used it and they quantified the relations between the operations of a manufacturing process, and the characteristics of quality generated by that process, using continuous variables with normal distribution. The results of this study form part of an initial phase of a project which seeks to explain the relationship among the different variables in a production process, which can be ordinal, discrete of continuous. Given that the SEM requires the variance and covariance matrix of the observed variables as access data, we have carried out a bibliographical check of the calculation of the correlations between the ordinal variables, and the results of this are presented in this article. They will later be used in the application stage in a real case in a production process.