Search

In this paper, we present an extension of the Poisson Zero-Modified model with Normal and Generalized Log-Gamma random effects. The random effect induces correlation and accommodate the intrinsic variability of each individual. The Generalized Log-Gamma effect is a generalized Normal effect and can be used in atypical situations where the Normal effect is not appropriate. In particular, the mixed Zero-Modified Poisson model allows us to deal with longitudinal count data, without requiring any previous knowledge about data characteristics, mainly to the number of zero observations (zero-inflated or zero-deflated). We consider the maximum likelihood approach to estimate the model parameters. A simulation study is presented to evaluate the estimators' performance. A real data set referring to the number of notification of infant deaths in the municipalities of the state of Bahia/Brazil is analyzed. The results revealed the Generalized Log-Gamma effect seems to be more appropriate to model this longitudinal data set. [ABSTRACT FROM AUTHOR]