Nonparametric hypothesis testing in a spatial-temporal model: A simulation study

Nonparametric procedures based on the bootstrap were developed for testing two assumptions in a spatial-temporal model, i.e., constant temporal effect across locations/spatial units and constant spatial effect over time. Simulation studies indicate that the procedures can correctly verify the assumptions for reasonably sized data. Presence of spatial clustering can further improve the sensitivity of the test under non-constant spatial effect over time. Furthermore, the test procedure for constant spatial effect over time is robust to model misspecification.