Limit theorems for random processes with random time substitution (Q1006998)

Let \(B[0,1]\) be the subspace of the Skorohod space \(D[0,1]\) consisting of nondecreasing, nonnegative functions and let \(\Pi[0,1]\subset B[0,1]\) be its subspace of continuous, strictly increasing functions \(\lambda\) with identical endpoint \(\lambda(1)\). If for every \(n\in\mathbb N\) independent random processes \(X_n\) and \(\Lambda_n\) are given such that convergence in distribution of either \(X_n\to X\) in \(D[0,\infty)\) and \(\Lambda_n\to\Lambda\) in \(\Pi[0,1]\) or \(X_n\to X\) in \(C[0,\infty)\) and \(X_n\to X\) in \(B[0,1]\) holds, then a functional limit theorem \(X_n\circ\Lambda_n\to X\circ\Lambda\) in \(D[0,1]\) for the composition of random processes is shown. This functional limit theorem is proved mainly by showing that the composition \(F(x,\lambda)=x\circ\lambda\) on \(D[0,\infty)\times D[0,1]\) endowed with the product Skorohod topology is continuous, when restricting considerations to the subspaces \(D[0,\infty)\times\Pi[0,1]\) or \(C[0,\infty)\times B[0,1]\). Continuity and the condition on identical endpoints in the definition of \(\Pi[0,1]\) are essential as shown by examples. The author further proves that under natural extra conditions the functional limit theorem can be extended to an almost sure version \[ \frac1{\log n}\sum_{k=1}^n\frac1k\,\delta_{X_n\circ\Lambda_n(\omega)}\to P_{X\circ\Lambda} \] almost surely, where \(\delta_x\) denotes Dirac measure in \(x\in\mathbb R\) and \(P_{X\circ\Lambda}\) is the distribution of the composed random process on \(D[0,1]\).