This paper presents a statistical model based on a multinomial distribution with fixed and random effects, the latter effects being structured and non-structured in space and time. Inference is performed through a Bayesian framework. We are interested in analyzing violent deaths at the level of parroquia in Ecuador. Noting that most of the deaths are linked to firearms, and much less with knifes, we build a multinomial model to predict the probability of three different types of deaths as a close proxy to a violent death having had occurred. We provide a practical and realistic interpretation of the model putting this in the real crime context and scenario in Ecuador.