摘要
对上海证券交易所综合股价指数（SSECI）收益率（5-min数据集）多标度条件下的相关性研究发现,在多重分形消除趋势波动法MF-DFA（适用于非稳定序列多点间相关性分析）的度量框架下,收益率序列G和｜G｜在10^1×5～10^5×5min的4个数量级标度条件下具有非唯一的幂相关指数,不同波动幅度表现出不同的长程相关性特征;收益率序列G的幂相关指数谱宽度大于序列｜G｜的指数谱宽度,平均谱指数小于｜G｜的平均谱指数,序列｜G｜表现出较G更强的正向持久性.
Systematic multiscale analysis of the power-correlation of SSECI was made on basis of 5-min database. The empirical findings show that under framework of multifractal detrended fluctuation analysis MF-DFA, return series G and absolute series I G I appears to exhibit the no-unique expo- nent of power-correlation over four magnitudes from 10^1 ×5 min to 10^5× 5 min. This exhibits that the different amplitudes have the different properties of long-range correlation. The power-correlation exponents of series G show the wider spectrum than for absolute return ｜G｜. The mean exponent of series ｜G｜ appears the stronger positive persistence compared to series G.
出处
《管理学报》
2007年第5期618-621,共4页
Chinese Journal of Management
基金
国家社会科学基金资助项目(05BJY010)
关键词
相关性
多标度
波动函数
幂相关指数
correlation
multiscale
fluctuation function
power correlation exponent
