RESUMO
Este artículo pretende mostrar, desde un punto de vista práctico, la manera adecuada de modelar una variable de respuesta de tipo recuento. Los resultados muestran que el modelo de regresión de Poisson es un modelo más adecuado que el de regresión lineal, ya que tiene en cuenta las características propias de los datos de recuento, aunque la presencia de sobredispersión indicará la corrección de los errores estándar de los parámetros del modelo de Poisson, y puede provocar el modelado mediante el modelo binomial negativa. El modelo de ceros aumentados será usado cuando haya un número excesivo de ceros. Para ello se modela el número de días de consumo de cannabis en función del consumo por parte del grupo de iguales y del consumo de cannabis del padre, de la madre y de los hermanos
This article aims to show the appropriate way to model a count type response variable from a practical viewpoint. The results show that the Poisson regression model is more appropriate than the linear regression model, as it takes count data characteristics into consideration, although the presence of overdispersion indicates the correction of standard errors in the Poisson model parameters and may cause modelling through the negative binomial model. The Zero Inflated model shall be used whenever there is an excessive number of zeros. To do so, the number of days of cannabis consumption in accordance with peer group consumption and cannabis consumption by parents and siblings is modelled
Assuntos
Masculino , Feminino , Adulto , Humanos , Fumar Maconha/epidemiologia , Abuso de Maconha/epidemiologia , Estudos Prospectivos , Modelos Lineares , Fumar Maconha/psicologia , Abuso de Maconha/psicologia , Análise MultivariadaRESUMO
The importance of adolescents taking part in various activities as a protective factor in substance consumption has been demonstrated over time. The objective of this paper is to analyze a sample of 1378 adolescents and the explanatory capacity of participating in different activities in relation to present-day alcohol, cannabis and tobacco consumption. Count variables were used as response variables, thus the Poisson Regression Model was applied in the analysis, within the context of a Generalized Linear Model. The results of the research demonstrate the explanatory value and differences in explanatory roles of each activity in connection with each the substances studied.
