CONSISTENT SPECIFICATION TESTING UNDER SPATIAL DEPENDENCE

We propose a series-based nonparametric specification test for regression function when data are spatially dependent, the “space” being of general economic or social nature. Dependence can be parametric, parametric with increasing dimension, semiparametric any combination thereof, thus covering vast variety settings. These include spatial error models varying types and levels complexity. Under new smooth dependence condition, our statistic is asymptotically standard normal. To prove latter property, we establish central limit theorem quadratic forms in linear processes an dimension setting. Finite sample performance investigated simulation study, bootstrap method also justified illustrated. Empirical examples illustrate real-world data.