Summary: Exploiting the classical R/S and modified R/S analysis, we first reveal the evidence of long-term memory in liquidity, volume, and volatility. Thereafter, we estimate the fractionally integrated autoregressive moving average ARFIMA models by both the exact-maximum likelihood (EML) and the modified-profile likelihood (MPL) methods. Furthermore, based on the theory of financial economics, we extend the simple ARFIMA models to the Multi-Factor ARFIMA models by incorporating the mutual relationships among financial market variables and present the effectiveness of the Multi-Factor ARFIMA models in financial markets.