The L1 and L2strong consistency of recursive kernel density estimators is established for mixing and ergodic stationary processes. Sharp rates of almost sure convergence are obtained in the L2 case for mixing processes. In addition, the notion and properties of Hilbert space-valued mixingales are developed, and strong laws of large numbers are given.
ASJC Scopus subject areas
- Information Systems
- Computer Science Applications
- Library and Information Sciences