The memory and heterogeneous effects existing on the financial time series have been widely revealed, especially in the crude oil market. While both effects, related to the temporal scales where two effects are measured, have been few studied. In this paper, we provide a comprehensive description of the dependence pattern of crude oil market by investigating the heterogeneity of autocorrelation of crude oil future in the framework of multiscale analysis and quantile regression analysis. Our empirical results are based on the price analysis of the West Texas Intermediate (WTI) crude oil future. We use quantile autoregression model to analyze the return and fluctuation series on multiple time scales. Firstly, we find that the autoregressive coefficients are likely to change with quantiles. Secondly, the autoregressive coefficients in the high-frequency components are small while in the low-frequency components are large. It indicates the fact that the crude oil price is a combination of random short-term fluctuation and deterministic long-term tendency. Interestingly, the quantile regressive coefficients for the return series are S-shaped while for the fluctuation series they are inverted S-shaped. In addition, the extreme lagged return and fluctuation are found to affect the distribution of the autoregressive coefficients.

金融时间序列上存在的记忆和异构影响已被广泛揭示，尤其是在原油市场中。虽然与影响两种效应的时间尺度有关的两种效应都鲜有研究。在本文中，我们通过在多尺度分析和分位数回归分析的框架下调查原油期货自相关的异质性，来全面描述原油市场的依赖模式。我们的经验结果基于西德克萨斯中质原油（WTI）原油期货的价格分析。我们使用分位数自回归模型来分析多个时间尺度上的收益和波动序列。首先，我们发现自回归系数可能随分位数而变化。第二，高频分量的自回归系数较小，而低频分量的自回归系数较大。这表明，原油价格是随机的短期波动和确定的长期趋势的组合。有趣的是，返回序列的分位数回归系数为S形，而对于波动序列的分位数回归系数则为S形。此外，发现极端滞后的回报和波动会影响自回归系数的分布。返回序列的分位数回归系数为S形，而波动序列的分位数回归系数为倒S形。此外，发现极端滞后的回报和波动会影响自回归系数的分布。返回序列的分位数回归系数为S形，而波动序列的分位数回归系数为倒S形。此外，发现极端滞后的回报和波动会影响自回归系数的分布。