A1158
Title: Measuring accordance movement between time series by Kendalls tau
Authors: Ying Zhang - Acadia University (Canada) [presenting]
Abstract: Choosing accordance/similarity measures between time series is critical in time series data science applications. There are different types of similarity measures, from traditional statistical correlation coefficients to measurements recently developed by computer scientists. Due to the lack of distribution properties, many of such measures have no statistical inference power. Motivated by an investigation of evaluating the performance of a gas multi-sensor device for monitoring urban pollution, we focus on measuring the accordance movement and its inference between a signal process to a reference process based on the method of Kendall Tau. Kendall Tau-based coefficients in measuring time series accordance movement are defined; it is shown how to make inferences by constructing conference intervals, and finally, the method with the pollution data from the gas multi-sensor device described above is demonstrated.