Detecting spatial clusters on functional data: a parametric scan statistic approach
share
This paper proposes a parametric scan statistic for detecting clusters of functional data indexed in space. The proposed method is based on an adaptation of a functional ANOVA. In a simulation study, it presents better performances than the nonparametric functional scan statistic for normal data and good performances also on quasi-normal distributions. The proposed method allows to detect smaller spatial clusters than the nonparametric one. It shows also better performances than the parametric univariate spatial scan statistic applied on the average over time. The parametric scan statistic for functional data was then applied on the search of spatial clusters of abnormal unemployment rates in France, considering the period from 1998 to 2013 divided in quarters.
READ FULL TEXT