Computational methods in the identification of forecasting time series models
-
Autores: África Ruiz Gándara, José María Caridad y Ocerín
- Localización: International journal of scientific management and tourism, ISSN-e 2386-8570, ISSN 2444-0299, Nº. 1, 2014, págs. 5-17
- Idioma: inglés
- Enlaces
-
Resumen
Most of the econometric software use goodness of fit measures as a tool to identify an ARIMA model, although the usual objective is forecasting. An alternative is to use predictive criteria for identification purposes. Both methods are compared with the aim of obtaining optimum forecasts. Some statistics such as BIC and others, obtained within the sample range are used together with some smoothing predictive criteria. They are compared as a tool for automatic selection of a useful model; the classical methods identify the correct model in just a third of the cases, as it is show with simulated series, and, in many cases a different model produces better forecasts. A case study with exports of the Spanish economy is presented with the comparative results attained with both approaches.
