Por favor, use este identificador para citar o enlazar este ítem:
http://cimat.repositorioinstitucional.mx/jspui/handle/1008/639| Forecasts by Predictive Likelihood in Threshold Autoregressive Model | |
| GRACIELA MARIA DE LOS DOLORES GONZALEZ FARIAS | |
| Acceso Abierto | |
| Atribución-NoComercial | |
| Predicción | |
| In this paper we study the performances of h-step ahead predictions in TAR mod- els versus those of AR models. We propose the alternative of Predictive Likelihood (PL), based on the principle of likelihood. Unlike other alternatives for prediction generation, PL jointly attacks the problems of estimating, obtaining and evaluting predictions. Which, for non linear models, presents an integrated way to take into account in both processes, the DGP. We also study the asymptotic properties of the maximum likelihood estimators under a known threshold parameter. We implement a wide simulation exercise to compare the predictions under PL and the standard predic- tion generating method, identified as the recursive Monte Carlo method. Comparisson of predictors performances is implemente d through the Mean Square Forecast Errors (MSFE). The gain observed in the performances of predictions by means of PL is clear and mainly due to the fact that the predictions for both the threshold variable and the autoregressive variable are obtained simultaneously. | |
| Centro de Investigación en Matemáticas AC | |
| 14-06-2007 | |
| Reporte | |
| Inglés | |
| Investigadores | |
| ANÁLISIS ESTADÍSTICO | |
| Versión publicada | |
| publishedVersion - Versión publicada | |
| Aparece en las colecciones: | Reportes Técnicos - Probabilidad y Estadística |
Cargar archivos:
| Fichero | Tamaño | Formato | |
|---|---|---|---|
| I-07-10.pdf | 404.02 kB | Adobe PDF | Visualizar/Abrir |