Use of Simulation Models for the Development of a Statistical Production Framework for Mobile Network Data with the simutils Package

Abstract: We propose to use agent-based simulation models for the development of statistical methods in Official Statistics, especially in relation with the new digital data sources. We present a mobile network data simulator which is managed through the simutils R package which provides geospatial representations of the simulated data. While the synthetic data are produced by an external tool, our simutils package allows an R user to parameterize and run this external simulation tool, to build geospatial data structures from the simulation output or to compute several aggregates. The geospatial data structures were designed with the purpose of using them in a visualization package too. Useful simulation models require the incorporation of real metadata from mobile telecommunication networks driving us to the inclusion of functionalities allowing the user to specify and validate them. All metadata are specified using XML file whose structure are defined in corresponding XSD files. Our R package includes example data sets and we show here how validate the metadata, how to run a simulation and how build the geospatial data structures and how to compute different aggregates.