Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
Multivariate Parametric Analysis of Interval Data
Paula Brito, A. Pedro Duarte Silva, José G. Dias

Last modified: 2013-06-14

Abstract


This work focuses on the study of interval data, i.e., when the variables’ values are intervals of IR, using parametric probabilistic models previously proposed. These models are based on the representation of each observed interval by its MidPoint and Midrange for which multivariate Normal and Skew-Normal distributions are assumed, considering different structures of the variance-covariance matrix. The proposed modelling has been applied to different multivariate methodologies- (M)ANOVA, discriminant analysis, model-based clustering - that are presented and discussed. The R-package MAINT. Data, available on CRAN, implements models and methods for the Gaussian case.

Full Text: PDF