Multivariate Bayesian spatio-temporal P-spline models to analyze crimes against women.

This article focuses on the development and application of multivariate spatio-temporal P-spline models to jointly analyze four forms of violence against women—rape, assault, cruelty by husband or relatives, and kidnapping and abduction—at the district level in Maharashtra, India, from 2001 to 2013. The proposed Bayesian models incorporate spatial and temporal smoothing via P-splines and account for correlations among different crimes to improve estimation precision and reveal potential associations between crime patterns. Model fitting is performed using integrated nested Laplace approximations (INLA) in R, addressing computational challenges typical of multivariate spatio-temporal models. The analysis identifies distinct spatial gradients and temporal trends among the crimes, with strong spatial and temporal correlations between rape, assault, and kidnapping, while cruelty exhibits a different pattern; these findings may inform further sociological investigation into underlying risk factors.