MORTALITY AMONG YOUNG NICARAGUAN IMMIGRANTS TO COSTA RICA: AN APPLICATION OF GEOGRAPHICALLY WEIGHTED STATISTICAL REGRESSION

Abstract

This paper applies a geographically weighted statistical regression (GWR) model to young Nicaraguan immigrant homicides in Costa Rica during the period 1998-2008 and identifies possible covariates. The pa- rameters for the GWR model are:Yi(g)=β0(g)+β1(g)x1 +β2(g)x2 +...+βk(g)xk +ε, which may be obtained from the solution ofβ(g) = (XT W (g)X)−1 XT W (g)Y.The GWR model is a more adequate model than the classic models, such as the log-linear Poisson model. In the GWR model, poverty was the most significant variable. The map of the estimators associated with the percentage of poor households suggests that the relationship between poverty and mortality by homicide for young Nicaraguan immigrants is stronger in the Caribbean region and neighboring zones. When the GWR model was run for homicides among young Costa Ricans, this effect was not observed, as it was among Nicaraguan immigrants.

Este trabajo aplica un modelo de regresión estadística especial ponderada (GWR) a los homicidios de inmigrantes nicaragüenses jóvenes en Costa Rica en el período 1998-2008 e identifica sus posibles covariables. Los parámetros del modelo GWRYi(g)=β0(g)+β1(g)x1 +β2(g)x2 +...+βk(g)xk +ε, β(g) = (XT W (g)X)−1 XT W (g)Y.El modelo GWR es un modelo más adecuado con respecto a modelos clásicos como el log-lineal de Poisson. En el modelo GWR la variable pobreza resultó la más significativa. El mapa de los estimadores asociados con el porcentaje de hogares pobres sugiere que la relación entre la pobreza y la mortalidad por homicidios de jóvenes nicaragüenses es más fuerte en el Caribe y zonas aledañas. Cuando el modelo GWR se aplicó a los homicidios entre jóvenes costarricenses, este efecto no se observó.