Una estimación robusta y no paramétrica de la transformación de Box y Cox para series de tiempo
Cargando...
Archivos
Autores
Calle Correa, Dasy Andrea
Director
Tipo de contenido
Document language:
Español
Fecha
Título de la revista
ISSN de la revista
Título del volumen
Documentos PDF
Resumen
En el análisis de series de tiempo estacionarias, es frecuente encontrarse que la varianza de la serie no es constante, siendo necesario en estos casos transformar la serie utilizando la familia de transformaciones introducida por Box y Cox (1964), donde se busca una transformación de potencia que permita estabilizar la varianza de la serie. Sin embargo, varios autores han estudiado y demostrado que la familia de transformaciones de Box y Cox parece no ser muy adecuada ni robusta cuando hay existencia de observaciones atípicas en la serie, la presencia de estas observaciones distorsiona la estimación del parámetro de transformación de Box y Cox. En este trabajo se plantea estudiar una propuesta para la estimación robusta del parámetro de la familia de transformaciones de Box y Cox que sea robusto ante la presencia de observaciones atípicas y tenga en cuenta el efecto de estas.
Abstract: In the stationary time series analysis it is frequent to encounter that the variance of the series is not constant. In these cases it is necessary to transform the series by using the family of transformations introduced by Box and Cox (1964), where it is sought a power transformation that allows stabilizing the variance of the series. However, several authors have studied and proved that the family of transformations Box and Cox seem not to be very adequate nor robust when there is presence of atypical observations in the series. The presence of these observations distorts the estimation of the parameter of transformation Box and Cox. In this work it is set out the study of a proposal to the robust estimation of the parameter of the family of transformations Box and Cox, which is robust facing the presence of atypical observations and takes in account the effect of these
Abstract: In the stationary time series analysis it is frequent to encounter that the variance of the series is not constant. In these cases it is necessary to transform the series by using the family of transformations introduced by Box and Cox (1964), where it is sought a power transformation that allows stabilizing the variance of the series. However, several authors have studied and proved that the family of transformations Box and Cox seem not to be very adequate nor robust when there is presence of atypical observations in the series. The presence of these observations distorts the estimation of the parameter of transformation Box and Cox. In this work it is set out the study of a proposal to the robust estimation of the parameter of the family of transformations Box and Cox, which is robust facing the presence of atypical observations and takes in account the effect of these