Crime Mapping and Spatial Data Analysis using R

This chapter introduced some basic principles of thematic maps. We learned how to make them using the tmap package, the importance of classification schemes, and how each one may produce a different looking map, which may tell a different story. For further reading, [Brewer (2006)] provides a brief condensed introduction to thematic mapping for epidemiologists but that can be generally extrapolated for crime mapping purposes. We have talked about the potential to develop misleading maps, and [Monmonier (1996)] "How to lie with maps" provides good guidance to avoid our negligent choices when producing a map confuse the readers. [Carr and Pickle (2010)] offers a more detailed treatment of small multiples and micromaps.

Aside from the issues we discussed around the computation and mapping of rates, there is a growing literature that is identifying the problem of ignoring that rates (as a standardising mechanism) using population in the denominator assumes that the relationship between crime and population is linear, when this is not generally the case. [Oliveira (2021)] discusses the problem and provides some solutions. Mapping rates, more generally, has been more thoroughly discussed within spatial epidemiology than in criminology. There is ample literature on disease mapping that address in more sophisticated ways some of the issues we introduce here (see [Waller and Gotway, 2004], [Lawson, 2021a], [Lawson, 2006], or [Lawson, 2021b]). Much of this work on spatial epidemiology adopts a Bayesian framework. We will talk a bit more about this later on, but if you want a friendly introduction to Bayesian data analysis we recommend [McElreath (2018)].