Density estimation for linear processes (Q802245)

Let \(X_ 1,...,X_ n\) be random variables forming a realization from a linear process \(X_ t=\sum^{\infty}_{r=0}g_ rZ_{t-r}\) where \(\{Z_ t\}\) is a sequence of independent and identically distributed random variables with \(E| Z_ 1|^{\epsilon}<\infty\) for some \(\epsilon >0\), and \(g_ r\to 0\) as \(r\to \infty\) at some specified rate. Let \(X_ 1\) have a probability density function f. It is then established that for every real x, the standard kernel type estimator \(\hat f_ n(x)\) based on \(X_ t\) (1\(\leq t\leq n)\) is, under some general regularity conditions, asymptotically normal and converges a.s. to f(x) as \(n\to \infty\).