Abstract: Studying the internal mechanism of jump and sorting out the different types of risk, is important for estimating and modeling volatility, which is the core content of risk management. Currently, based on high frequency data, the research in this field is still in its infancy, so there are quite rich contents looking forward to be explored. Based on non-parametric approach, the new jump variance and continuous sample path variance are constructed and jump variance is modeled by combining A-J jump detection statistic. With high frequency data from Shanghai composite index, the empirical analyses are carried out. It turns out that the jump variance show leptokurtic, heavy tail and volatility clusters; the contribution of jump variance to whole variance nearly equals for different sampling frequency; the positive jump and negative jump are asymmetry and the adjusted returns are nearly normal distribution; the correlation between jumps and economic information release is always positive. Further,some abnormal phenomenas are explained. This study can be applied to help investors optimize strategy of investment and provide regulatory basis for the regulatory authorities, according to complexity of the volatility and jump.

Key words: high frequency data, jump, volatility modeling