Modelo de regresión lineal múltiple con errores distribuidos secante hiperbólica generalizada
Cargando...
Archivos
Autores
Director
Tipo de contenido
Document language:
Español
Fecha
Título de la revista
ISSN de la revista
Título del volumen
Documentos PDF
Resumen
En este trabajo se presenta un estudio del modelo de regresión lineal múltiple, donde el error tiene distribución secante hiperbólica generalizada. Para estimar los parámetros del modelo, se utiliza la Máxima Verosimilitud Modificada (MVM), que permite linealizar las formas funcionales que no son lineales y obtener soluciones explicitas cerradas. Además, se realiza todo el estudio formal de las propiedades asintóticas e inferencia de los estimadores obtenidos, mediante el uso de la simulación de Monte Carlo, que permite comprobar aspectos como: los estimadores de MVM son más eficientes y robustos que los de estimadores de Mínimos Cuadrados (MC). Mediante varios ejemplos de la literatura estadística, se valida la metodología desarrollada y descrita en este trabajo.
Abstract: In this work, a study of multiple linear regression is presented, where the error has a distribution secant hyperbolic generalized. Furthermore, to estimate the parameters of this model, is used the Modified Maximum Likelihood (MML), which enables linearize the functional forms because it is not linear, and obtains solutions in a closed form. In addition, the formal studies of asymptotic properties and inference of these estimators obtained, using the Monte Carlo simulation, which allows to check aspects such as: MML estimators are more efficient and robust than those of least squares estimators (LS). Through several examples of statistical literature, the methodology developed and described in this work is validated.
Abstract: In this work, a study of multiple linear regression is presented, where the error has a distribution secant hyperbolic generalized. Furthermore, to estimate the parameters of this model, is used the Modified Maximum Likelihood (MML), which enables linearize the functional forms because it is not linear, and obtains solutions in a closed form. In addition, the formal studies of asymptotic properties and inference of these estimators obtained, using the Monte Carlo simulation, which allows to check aspects such as: MML estimators are more efficient and robust than those of least squares estimators (LS). Through several examples of statistical literature, the methodology developed and described in this work is validated.