数据预处理对替代数据检验方法的影响

替代数据检验法是检验时间序列中是否存在确定性非线性成分的重要统计方法.通过研究差分和数据平滑运算对替代数据检验方法的影响,指出常用的线性滤波等数据预处理步骤破坏了序列的静态性质,从而会导致对零假设的错误拒绝.因此,建议应该直接利用原始时间序列而非应用了差分等非静态滤波运算后的时间序列生成替代数据,再进行假设检验,以免造成对零假设的错误拒绝.

Effect of the Data Processing for Surrogate Data Tests

In the analysis of time series,the surrogate data test is often performed in order to investigate nonlinearity in the data.When we use this method,we find that the test does not always succeed to reject the null hypotheses when we apply the differencing or smoothing operator to the original data sets.Because it is not a static nonlinear transform of a linear process.So it is suggested that one should creating the surrogate data sets directly from the original time series,otherwise spurious results can occur.