ARFIMA models are characterized by both their long-range dependence and fractional values for the ARIMA model differencing parameter. Stationarity is achieved for d <FONT FACE=Symbol>Î</FONT> (-0.5, 0.5) and the long memory appears whether d is positive. This work tests and compares two methodologies for the differencing parameter estimation based on, respectively, Periodogram and Smoothed Periodogram functions. Through synthetic series generated to this purpose, simulations were ran to four different ARFIMA structures: (0,d,0), (1,d,0), (0,d,1), (1,d,1) and three values of d (0,0; 0,10; 0,25 and 0,40).
ARFIMA; Long Memory; Fractional d; Periodogram