Conference is now cancelled

A (corrected) graphical hypothesis test of spatiotemporal clustering and clustering range estimation for the tau statistic.

Timothy M Pollington (MathSys, University of Warwick; visiting Big Data Institute, University of Oxford.)

The tau statistic τ uses geolocation and symptom onset time to assess global spatiotemporal clustering and estimate its extent, for epidemiological studies [1,2]. We explore how implementation could bias τ estimates [4]. Graphical hypothesis tests or parameter estimation for τ have been applied incorrectly in the past [3]; I explain how Diggle highlighted this error and suggested the correction. Using the Hagelloch open-access measles dataset, we compare a previous analysis as the baseline [2] vs. our modified implementation of a graphical hypothesis test, and estimated clustering range by i) bootstrap sampling method; ii) number of bootstrap samples N; iii) CI type (bias-corrected & accelerated (BCa) vs percentile); and iv) distance band sets.

[1] Salje et al. PNAS. 2012 [2] Lessler et al. PLoS One. 2016 [3] Pollington et al. arXiv/stat.AP. 2019 [4] ibid. arXiv/stat.ME. 2019