Modelos lineales generalizados para el análisis de defunciones en menores de un año en la región Sierra del Ecuador.

Abstract

The goal of this study was to characterize and model deaths in children under one year of age, using generalized linear models. With the descriptive and exploratory statistics, two (2) quantitative variables (Deaths and Births) and three (3) nominal qualitative variables (Gender, Causes of Death, Provinces) were identified as relevant. It was also obtained that, at the provincial level, the behavior of the mortality rate is not homogeneous. It was observed strong differences between the entities in terms of mortality during the first year of life. The province of Bolívar is the one that leads. However, there is a considerable number of deaths in men in Pichincha province. It was concluded that the main cause of death is prenatal period conditions. Using simple regression techniques, the factors that have contributed to infant mortality were identified, as well as the variables that explain the interprovincial differences of this indicator except for the gender. In the same way, a multiple linear regression analysis was made getting an Akaike Information Criterion (AIC) of 427.66 and a R^2 of 0.9666, However, the model did not meet the assumptions. For this reason, it was not suitable to predict. Therefore, generalized linear models were applied, especially Poisson regression, getting an outstanding adjustment, a R^2 of 858 and an AIC of 650.72. Finally, comparative simulations were carried out with the purpose of evaluating the possibility of applying the same methodology to mortality data from all over the country, and it was concluded that it is possible.