Modeling a proportion variable in a complex survey through the Beta distribution: The case taken from the CASEN survey