Um modelo robusto assimétrico de análise fatorial

Abstract

In this work we develop an extension of the classic factor analysis model, by relaxing the assumption of normality of the factors. Instead, we suppose that the joint distribution of the factors and observational errors is a scale mixture of skew-normal distributions, allowing us to model data following a nonstandard pattern, presenting skewness and heavy tails at the same time. A relevant feature of the model is the parametrization used for the scale mixture, defined in such a way that all the elements of the shape vector but the first are guaranteed to be zero.