Una extension de la regresion propuesta por Geweke y Porter-Hudak para la estimacion del orden de diferenciacion en modelos ARFIMA

Over the last years there has been considerable interest in the application of long memory time series models in economics using ARFIMA models. Nowadays, the most popular estimator of the difference parameter in economic applications is that proposed by Geweke and Porter-Hudak (GPH) although has been shown that, in particular circumstances, this estimator can be badly biased. So we propose a biased-corrected extension of that estimator using the exponential model proposed by Bloomfield. The performance of both estimators are investigated for moderate-size samples and it is concluded that the extension proposed has smaller mean squared error than the GPH estimator.