Measurability problems for empirical processes (Q1091031): Difference between revisions

Let \(X: T\times \Omega \to {\mathbb{R}}^ n\) be a uniformly bounded stochastic process, defined on a complete probability space (\(\Omega\),\(\Sigma\),P) with abstract index set T. The author introduces the notion of a consistent lifting \(\rho\) of \(L^{\infty}(\Omega)\) in such a way that in general the process \(\rho\) X has much better measurability properties than X and that \(\bar X=| \rho X-X|\) satisfies \(\bar X{}_ t=0\) a.s. for each \(t\in T.\) It is shown that X satisfies the CLT if and only if \(\rho\) X and \(\bar X\) satisfy the CLT, meaning that convergence in distribution (in an abstract sense) of appropriately centered and normalized partial sums to some Radon measure holds. This result is then applied to derive the following contraction principle without any measurability assumptions: Let X,Y be two uniformly bounded processes with \(| Y_ s(\omega)- Y_ t(\omega)| \leq | X_ s(\omega)-X_ t(\omega)|\) for all \(\omega\in \Omega\) and \(t\in T\). If X satisfies the CLT, then Y also satisfies the CLT.