Sesia Ilaria, Cantoni Elena, Cernigliaro Alice, Signorile Giovanna, Fantino Gianluca, Tavella Patrizia
IEEE Trans Ultrason Ferroelectr Freq Control. 2016 Apr;63(4):575-81. doi: 10.1109/TUFFC.2015.2496280. Epub 2015 Oct 30.
The Allan variance (AVAR) is widely used to measure the stability of experimental time series. Specifically, AVAR is commonly used in space applications such as monitoring the clocks of the global navigation satellite systems (GNSSs). In these applications, the experimental data present some peculiar aspects which are not generally encountered when the measurements are carried out in a laboratory. Space clocks' data can in fact present outliers, jumps, and missing values, which corrupt the clock characterization. Therefore, an efficient preprocessing is fundamental to ensure a proper data analysis and improve the stability estimation performed with the AVAR or other similar variances. In this work, we propose a preprocessing algorithm and its implementation in a robust software code (in MATLAB language) able to deal with time series of experimental data affected by nonstationarities and missing data; our method is properly detecting and removing anomalous behaviors, hence making the subsequent stability analysis more reliable.
阿仑方差（AVAR）被广泛用于测量实验时间序列的稳定性。具体而言，AVAR常用于空间应用，如监测全球导航卫星系统（GNSS）的时钟。在这些应用中，实验数据呈现出一些特殊的方面，这些方面在实验室进行测量时通常不会遇到。实际上，空间时钟的数据可能存在异常值、跳变和缺失值，这会破坏时钟的特性描述。因此，有效的预处理对于确保正确的数据分析以及改进使用AVAR或其他类似方差进行的稳定性估计至关重要。在这项工作中，我们提出了一种预处理算法及其在健壮软件代码（用MATLAB语言编写）中的实现，该代码能够处理受非平稳性和缺失数据影响的实验数据时间序列；我们的方法能够正确检测并去除异常行为，从而使后续的稳定性分析更加可靠。
