A Comparative Study for Estimate Fractional Parameter of ARFIMA Model

Long memory analysis is one of the most active areas in econometrics and time series where various methods have been introduced to identify estimate long parameter partially integrated series. One common models used represent that a ARFIMA (Auto Regressive Fractional Integration Moving Average Model) which diffs are fractional number called parameter. To analyze determine model, fractal must be estimated. There many for estimation. In this research, estimation were divided into indirect methods, Hurst estimated first, then integration from it by relation between them. As direct directly without relying on Hurst's parameter, them semi parametric methods. paper, some fraction modulus namely (Geweke-Porter-Hudak, Smoothed Geweke-Porter-Hudak, Local Whittle, Wavelet weighted wavelet), using simulation method with different value (d) size The comparison was done mean squared error (MSE). It turns out best (Local Whittle).
 model generated function programmed MATLAB statistical program