An important issue in functional time series analysis is whether an observed series comes from a purely random process. We extend the BDS test, a widely-used nonlinear independence test, to the functional time series. Like the BDS test in the univariate case, the functional BDS test can act as the model specification test to evaluate the adequacy of various prediction models and as a nonlinearity test to detect the existence of nonlinear structures in a functional time series after removing the linear structure exhibited. We show that the test statistic from the functional BDS test has the same asymptotic properties as those in the univariate case and provides the recommended range of its hyperparameters. Additionally, empirical data analysis features its applications in evaluating the adequacy of the fAR(1) and fGARCH(1,1) models in fitting the daily curves of cumulative intraday returns (CIDR) of the VIX index. We showed that the functional BDS test remedies the weakness of the existing independence test in the literature, as the latter is restricted in detecting linear structures, thus, can neglect nonlinear temporal structures.
翻译:功能时间序列分析中的一个重要问题是观察到的序列是否来自一个纯随机过程。我们将BDS测试,一种广泛使用的非线性独立性测试,扩展到功能时间序列。与单变量情况下的BDS测试一样,功能BDS测试可以作为模型规范测试,以评估各种预测模型的适用性,并作为非线性测试,以在去除线性结构之后检测功能时间序列中非线性结构的存在。我们证明了来自功能BDS测试的测试统计量具有与单变量情况下相同的渐近特性,并提供其超参数的推荐范围。此外,实证数据分析展示了其在评估fAR(1)和fGARCH(1,1)模型拟合VIX指数稀疏交易的每日累计收益曲线(CIDR)方面的应用。我们展示了功能BDS测试纠正了文献中现有独立性测试的缺陷,后者受限于检测线性结构,因此可能忽略非线性时间结构。