Evaluating the Geographical Patterns of Intimate Partner Violence in Nigeria, Zambia, and Zimbabwe Using INLA-SPDE Approach

Recent data show that one in three women in Sub-Saharan Africa is abused at least once in her lifetime by an intimate partner. The fight against violence against women has thus been included among the targets of two Sustainable Development Goals of the United Nations. The literature shows that there exists a complex interplay between personal and contextual characteristics, and where community-level characteristics such as living in rural areas come into play, there is a higher presence of intimate partner violence. The aim of this preliminary study is to examine the geographical patterns of the number of victims of intimate partner violence in Nigeria, Zambia, and Zimbabwe, using the Demographic and Health Survey (DHS) data. Indeed, this survey provides data on current and past experiences of abuse and the geolocation of its clusters. These data were used in a Bayesian hierarchical modelling approach within the framework of INLA-SPDE, to model the total number of victims at the cluster level – which is assumed to be continuously observed in an underlying Gaussian random field and with a Poisson distribution. The models were explored with one final model selected for each country, and posterior estimates of the model parameter values (along with the 95% credible intervals) were analysed. Preliminary results show that people living in rural areas are more exposed to violence. They also provide a useful evidence base for policymakers to create programs centred around the people living in the identified hotspots whilst taking specific cultural practices of their residents into consideration.