Based on Long Memory Research of Copper Index Fluctuation with Introducing Positions

摘要

This paper introduces Co-integration test, Modified model for error, Vector auto-regression, Granger causality test, Impulse response function, and proves the necessity in the process of establishment for models with the sequence of positions. Using of modified R/S methods, we have built up the ARFIMA models, FIGARCH models, ARFIMA-FIGARCH models for the fluctuation of copper index earnings, and have given the analysis of the sequence of earnings r_t, the sequence of fluctuation for earnings |r_t |, the sequence of residual |ε_t |. Furthermore, we can make more precise forecast with the models ARFIMA(0, d_1,0) -FIGARCH(1, d_2,1).