Crime patterns in Delhi: a Bayesian spatio-temporal assessment

The present research is aimed at crime modelling in the union territory of Delhi using Bayesian spatio-temporal methodology. Objective of conducting the present study is to identify crime risk and crime propensity for property crime incidents, based on the compounded data-prior belief approach, in the districts of the union territory of Delhi. Appropriate subjective priors are formulated using to represent both vague and available information from the geographically adjacent neighboring units. Suitability of intrinsic conditional autoregressive prior for investigating spatially structured random effects ( $$s_{i}$$ s i ) and spatio-temporal interaction term ( $$\delta_{i}$$ δ i ) for crime counts is justified through caterpillar plots. Crime mapping using Bayesian tool kit helps in demarcation of hot and cold spots. Northern and Eastern regions of Delhi are found to have more crime propensity. Posterior predictive analysis reaffirms choice of Poisson model for district wise crime patterns. Areal literacy rate and proportion of non-workers are established as significant influencers in occurrence of crime. Superiority of Bayesian spatio-temporal methods over the classical methodologies as well as over the classical spatio-temporal studies on crime is asserted.