Journal of Basic and Applied Sciences - Volume 8 Number 2
Poisson Regression Models for Count Data: Use in the Number of Deaths in the Santo Angelo (Brazil)
Suzana Russo, Diego Flender and Gabriel Francisco da Silva
http://dx.doi.org/10.6000/1927-5129.2012.08.02.01
Abstract: When speaking about data, presuppose good quality to these. Otherwise would be affected the accuracy of the information, which would lead to false interpretations. In Health Statistics data are obtained through surveys, presented in its simplest expression, taking advantage of existing records, making an inquiry, or by means of experiments. The rational organization of the data allows characterizing what are the priority issues and thus establishing health programs. To analyze the mortality data it is necessary to be considered the mortality in certain age groups, so we can find data that show the incidence of major groups of deaths. From the analysis of data is followed by subsequent formulation of the Poisson regression models, where each group in question, by age group is represented by a number of counting time. The Poisson regression model is a specific type of Generalized Linear Models (GLM) and non-linear. As [1], its main features are: a) provides, in general, a satisfactory description of experimental data whose variance is proportional to the mean, b) can be deduced theoretically from first principles with a minimum of restrictions; c) If events occur independently and randomly in time with constant average rate of occurrence, the model determines the number of time specified. At the end of this study could be seen that through the analysis of the data could be found that the age group from 70 to 79 years old is the highest incidence of deaths with 21.1%, then comes the range of 60 to 69 years old with 20%, and was also recorded for time worked, in January 2000 to December 2004, the death rate was 52.27, and variance equal to 102.43, the city of Santo Angelo (Brazil). It was further found that the data are analyzed overdispersion, variance greater than average, then it was necessary to remove the overdispersion to find the appropriate template. With the pattern found was made short-term forecasts.
Keywords: Deaths, Poisson regression models, Overdispersion
