Modelos de regressão aditivos dinâmicos aplicados a dados logitudinais

Abstract

This work focuses on the regression additive dynamic models for longitudinal data, suggested in the versions of Borgan et al. (2007) and Martinussen and Scheike (1999, 2000). The method is intended for recurrent event data of interest in longitudinal studies that involves multiple analysis units. In both versions is modeled conditional meangiven the previous history in terms of fixed and time-dependent covariates. Based on the asymptotic results are presented the confidence intervals, the confidence bands and the hypothesis testing. The goodness of fit is considered through the use of martingales residuals.Two applications are presented, each one is referring to the methodologies studied. The data comes from the studies of Serrinha's city conducted by the Public Health Institute, Federal University of Bahia. In each application different parts of the data are used, but with the same purpose of evaluating the effect of periodic supplementation from vitamin A about the diarrhea in children under 5 years old.