Comparable Analysis of Long-term Memory of EUR/USD Based on Non-parametrical Statistics

Abstract

Non-parameter statistical methods of classical R/S, revised R/S and V/S was proposed in long-term Memory of EUR/USD. It was concluded through comparable analysis that: (1) Through Normality Test of Return Sequences of EUR/USD Daily Closing Prices, the experimental results imply that bias of daily return sequences of EUR/USD is not equal to zero, and the curve appears to high peak and fat tail. (2) Jarque-Bera test is adopted. The estimated values reject the null hypothesis of normal distribution. (3) The paper estimates daily return series of EUR/USD using classical R/S method. The results show that Hurst exponent is equal to 0.612425; the statistical cycle is 160 days; the correlative scale is close to 1.3432. This study's conclusion was that long-term memory exists in daily return time series of EUR/USD is proved.Key words: rescaled range (R/S) analysis; rescaled variance (V/S) analysis; long-term memory