[GIS] How to Construct Spatio Temporal Clusters of Time Series Data in R

clusteringhot-spotr

I have count data of one disease for various districts of a state of India on monthly basis. Is it possible to draw spatio temporal clusters in R? I have checked some R- packages for example surveillance and got a paper which states a model 'hhh4' in Section 5 which nearly talks about my problem. But I am not good at data objects of various geographical classes. Is there any way from which I can use my csv file (which contains months, case count and lon lat) and analyse significant clusters over time (in R, secondarily ArcGis)?

Best Answer

The "endemic-epidemic" modelling framework hhh4 in surveillance focusses on parameter inference (e.g., epidemic proportion) and short-term forecasts for (multivariate) count time series, see the corresponding vignette("hhh4_spacetime"). The package does contain tools for outbreak detection but these are limited to univariate count time series and categorical time series, see the vignette("monitoringCounts").

For space-time anomaly detection in R you have to resort to another package:

scanstatistics - https://CRAN.R-project.org/package=scanstatistics

It contains an introductory vignette, which explains and exemplifies how to apply different scan statistics based on a dataset of counts by region and time period and a dataset of corresponding spatial coordinates, such as provided by your csv file.

Related Question