Application of Longitudinal Analysis to Crime Data: Windhoek Case study (2011-2016)

Abstract

Crime in the Windhoek municipal area continue to be on an increase trend as the population grow over time. Despite past effort done to reduce crime, crime seems to be on a continuously increasing trend; mostly in area regarded as crime hotspots by Windhoek municipal Police. Past study done to analyse the crime record have concentrated mostly on cross-sectional analysis,
which does not take correlation into account, thus makes it difficult to compare snapshots of crime over time. The main aim of this research was to analyse reported crime data for the period (2011-2016), using a more robust method known as longitudinal data analysis. This method helped us to visualise the different crime frequencies at different time points (month, day or time of day) in all identified police zones. Furthermore, the use of Generalised Estimating Equations (GEE) was also done, to model these crime data, where the best correlation structure was identified to be the exchangeable correlation structure, which assume constant correlation over time.