Estimating Latent Linear Correlations from Fuzzy Frequency Tables

Abstract This research concerns the estimation of latent linear or polychoric correlations from fuzzy frequency tables. Fuzzy counts are particular interest to many disciplines including social and behavioral sciences especially relevant when observed data classified using categories—as for socioeconomic studies, clinical evaluations, content analysis, inter-rater reliability analysis—or imprecise observations into either precise analysis ratings fuzzy-coded variables. In these cases, space count matrices is no longer defined over naturals and, consequently, estimator cannot be used accurately estimate correlations. The aim this contribution twofold. First, we illustrate a computational procedure based on generalized natural numbers computing frequencies. Second, reformulate problem estimating in context expectation–maximization-based maximum likelihood estimation. A simulation study two applications investigate characteristics proposed method. Overall, results show that EM-based more efficient deal with as opposed standard estimators may context.