Kernel-type estimators of the multivariate density of stationary random fields indexed by multidimensional lattice points space are investigated. Sufficient conditions for kernel estimators to converge uniformly are obtained. The estimators can attain the optimal rates L∞ of convergence. The results apply to a large class of spatial processes.
Relation:
Statistics & Probability Letters, Volume 36, Issue 2, 1 December 1997, Pages 115-125