The empirical process of a short-range dependent stationary sequence under Gaussian subordination (Q1908536)

Consider the stationary sequence \(X_1 = G(Z_1)\), \(X_2 = G(Z_2),\dots\), where \(G (\cdot)\) is an arbitrary Borel function and \(Z_1, Z_2, \dots\) is a mean-zero stationary Gaussian sequence with covariance function \(r(k) = E(Z_1 Z_{k + 1})\) satisfying \(r(0) = 1\) and \(\sum^\infty_{k = 1} |r(k) |^m < \infty\), where, with \(I \{\cdot\}\) denoting the indicator function and \(F (\cdot)\) the continuous marginal distribution function of the sequence \(\{X_n\}\), the integer \(m\) is the Hermite rank of the family \(\{I \{G (\cdot) \leq x\} - F(x) : x \in \mathbb{R}\}\). Let \(F_n (\cdot)\) be the empirical distribution function of \(X_1, \dots, X_n\). We prove that, as \(n \to \infty\), the empirical process \(n^{1/2} \{F_n (\cdot) - F (\cdot)\}\) converges in distribution to a Gaussian process in the space \({\mathcal D} [- \infty, \infty]\).