Métodos de regresión para Count Data
Fecha
2017-06
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Jaén: Universidad de Jaén
Resumen
El presente “TFG” está enfocado en describir detalladamente el modelado estadístico sobre modelos específicos de regresión para count data (datos de conteo o recuento), dado que los modelos de regresión clásicos no son adecuados, ya que presentan claras deficiencias cuando la variable dependiente es limitada, como ocurre con este tipo de datos de uso habitual; siendo una alternativa los modelos de elección discreta. Se realiza un breve resumen teórico que nos permite conocer a priori las medidas más utilizadas para poder trabajar con estos tipos de modelos, con el propósito de aplicarlas para analizar el número de víctimas de accidentes de tráfico que se registran en los puntos negros de la red vial española mediante el lenguaje de R. Esta variable, por definición, es considerada una variable truncada, ya que toma valores a partir del valor 3. Por este motivo, además de utilizar el modelo de Poisson y el modelo binomial negativo, entre otros, como modelos específicos para analizar datos de recuentos, utilizamos modelos truncados; siendo una extensión de los mismos.
The present "Final Degree Project" is focused on describing in detail the statistical model about specific models of regression for count data, since the classic regression models are not adequate, since they present clear deficiencies when the dependent variable is limited, as occurs with this type of data of habitual use; being an alternative discreet choice models. A brief theoretical summary is made, which allows us to know a priori the most used measures to be able to work with these types of models, with the purpose of applying them to analyze the number of victims of traffic accidents that are registered in the black points of the spanish roads through the language of R. This variable, by definition, is considered a truncated variable, since it takes values from the value 3. For this reason, in addition to using the Poisson model and the negative binomial model, among others, as specific models for analyzing count data, we use truncated models; being an extension thereof.
The present "Final Degree Project" is focused on describing in detail the statistical model about specific models of regression for count data, since the classic regression models are not adequate, since they present clear deficiencies when the dependent variable is limited, as occurs with this type of data of habitual use; being an alternative discreet choice models. A brief theoretical summary is made, which allows us to know a priori the most used measures to be able to work with these types of models, with the purpose of applying them to analyze the number of victims of traffic accidents that are registered in the black points of the spanish roads through the language of R. This variable, by definition, is considered a truncated variable, since it takes values from the value 3. For this reason, in addition to using the Poisson model and the negative binomial model, among others, as specific models for analyzing count data, we use truncated models; being an extension thereof.