Spatial Item Factor Analysis With Application to Mapping Food Insecurity
share
Item factor analysis is widely used for studying the relationship between a latent construct and a set of observed variables. One of the main assumptions of this method is that the latent construct or factor is independent between subjects, which might not be adequate in certain contexts. In the study of food insecurity, for example, this is likely not true due to a close relationship with socio-economic characteristics, that are spatially structured. In order to capture these effects, we propose an extension of item factor analysis to the spatial domain that is able to predict the latent factors at unobserved spatial locations. We develop a Bayesian sampling scheme for providing inference and illustrate the explanatory strength of our model by application to a study of the latent construct `food insecurity' in a remote urban centre in the Brazilian Amazon. We use our method to map the dimensions of food insecurity in this area and identify the most severely affected areas. Our methods are implemented in an R package, spifa, available from Github.
READ FULL TEXT
