A Skorohod representation theorem for uniform distance (Q718887)

Let \(D\) be the set of real càdlàg functions on \([0,1]\), \(d\) the Skorohod distance and \({\mathcal B}_d\) the Borel \(\sigma\)-field on \({\mathcal B}_d\). Skorohod's representation theorem says that, if \((\mu_n)\) is a sequence of probability measures that converges weakly to the probability measure \(\mu_0\), then there exist \(D\)-valued random variables \(X_n\) and \(X\) such that \(d(X_n,X_0)\to 0\) and, for every \(n\geq 0\), \(\mu_n\) is the law of \(X_n\). However, \(X_n\) may fail to approximate \(X_0\) uniformly. The authors solve the following problem: Find necessary and sufficient conditions for the validity of the statement: ``On some probability space \((\Omega,{\mathcal F},P)\), there exist \(D\)-valued random variables \(X_n\) \((n\geq 0)\) such that, for every \(n\geq 0\), \(\mu_n\) is the law of \(X_n\) and \(\|X_n-X_0\|\) converges to \(0\) in probability''. Here, \(\|\cdot\|\) is the supremum norm. They prove (Theorem 4) that the necessary and sufficient condition is \[ \lim_{n\to+\infty}\,\sup_{f\in L} |\mu_n(f)-\mu_0(f)|=0, \] where \(L\) is the set of functions \(f:D\to\mathbb R\) satisfying \[ \sigma(f)\subset\mathcal{B}_d,\quad -1\leq f\leq 1, \quad |f(x)-f(y)|\leq\| x-y\|\quad (x,y\in D). \] They also apply this result to exchangeable empirical processes and to pure jump processes.