A multivariate Gaussian scan statistic forspatial data

Spatial autoregressive models for scan statistic

“…Recently, Cucala et al (2017) designed a Gaussian scan statistic for spatial multivariate data. Their method takes into account the correlation between different variables but assumes independence between neighbouring sites.…”

Investigating spatial scan statistics for multivariate functional data

“…Then we chose to obtain a large set of simulated datasets by randomly permuting the observations X i in the spatial locations. This technique called "random labelling" was already used in spatial scan statistics (Kulldorff et al, 2009;Cucala et al, 2017;Frévent et al, 2021).…”

“…For this purpose three different approaches can be considered. The simplest one consists in summarizing the information by averaging each variable over the time and to apply a parametric multivariate spatial scan statistic (Cucala et al, 2017) or the nonparametric one proposed by Cucala et al (2018) but this could lead to a huge loss of information when the data is measured over a long time period. Another solution could be to apply a spatial scan statistic for univariate functional data on each variable (Smida et al, 2020;Frévent et al, 2021).…”

An up-to-date review of scan statistics

“…It is important to mention that scan statistics can be extended to deal with different kinds of data, for example event data (Rosychuk, Huston and Prasad, 2006), ordinal data (Jung, Kulldorff and Klassen, 2007), continuous data (Kulldorff, Huang and Konty, 2009;Zhang, Zhang and Lin, 2012), multivariate data (Cucala et al, 2017), and others. Zhang and Lin (2014) noticed that the likelihood ratio statistic is a special case in the family of power divergence (PD) goodness-of-fit statistics, so the classical spatial scan test can be extended to the family of PD spatial scan tests.…”

“…Sometimes it is interesting to consider the correlation between variables in spatial clustering problems. Cucala et al (2017) applied a multivariate Gaussian scan statistic to find spatial clusters. This method is more powerful than its independent version.…”