Cluster-weighted disjoint factor analyzers for exploring the impact of socioeconomic factors on crime rates.

Motivated by the analysis of the crime heterogeneity in the communities within the United States, we investigated the effects of socio-economic information of the communities on crime rates. We aim to identify sub-groups of communities hidden within the United States with homogeneous effects of socio-economic information on crime rates. Moreover, we also identify disjoint groups of socio-economic features that similarly predict crimes within each community group. Identifying the homogeneous communities in terms of crimes is particularly important since this would help policymakers choose cluster-specific policy inter- ventions in those areas. To achieve this, we employ the Multivariate Cluster-Weighted Disjoint Factor Analyzers (MCWDFA), enabling us to i) cluster communities based on their socio-economic features on crime rates; (ii) identify cluster-specific sub-groups of socio-economic charac- teristics with similar effects on the selected crime rates. Results confirm significant heterogeneity in crimes across United States communities and diverse effects of socio-economic information on crime rates within each community group.