Desarrollos recientes en estadística: aportes teórico-metodológicos a la investigación sociológica

Abstract

Current manuscript shows and discuss the empirical advantages of employing modern statistical tools (i.e., polychoric correlations, item factor analysis and structural equation modeling) specifically developed to analyze latent variables measured with ordinal categorical items, and compared these techniques to two classic statistical analyses developed for observed (non-latent) variables measured at an interval level of measurement (i.e., Pearson correlations and linear regression analysis). Results from two Monte Carlo studies revealed that statistical methods specifically developed to work with categorical observed variables and latent variables allowed for better characterization of measurement instruments and theoretical constructs, as well as more precise characterization of the strength of the association between variables and the overall variance explained by the model. The relevance of modern statistical techniques for sociological quantitative research is discussed in order to enlighten a possible shift in the methodological and epistemological scope in the social sciences from simple null-hypothesis testing to the testing of alternative theoretical models.