Spatial kernel regression estimation: weak consistency.

Abstract

In this paper, we introduce a kernel method to estimate a spatial conditional regression under mixing spatial processes. Some preliminary statistical properties including weak consistency and convergence rates are investigated. The sufficient conditions on mixing coefficients and the bandwidth are established to ensure distribution-free weak consistency, which requires no assumption on the regressor and allows the mixing coefficients decreasing to zero slowly. However, to achieve an optimal convergence rate, some requirements on the regressor and the decreasing rate of mixing coefficients tending to zero are needed.