Smoothing metrics for measures on groups (Q912461)

Let G be a locally compact Abelian group with countable basis. A real valued continuous and integrable function g on G vanishing at infinity (i.e. \(g\in C_ 0(G)\cap {\mathcal L}^ 1(G))\) is called a metric kernel if the zero set Z(g) of its Fourier transform has void interior. In fact, there exist metric kernels, and \(g\in C_ 0(G)\cap {\mathcal L}^ 1(G)\) is a metric kernel iff its translates \(g_ x\), \(x\in G\), are total in \((C_ 0(G),\|\) \(\|_{\infty})\) (Theorem 2.5). Then by \(\| \mu \|_ g=\| g*\mu \|_{\infty}\) there is defined a norm \(\| \cdot \|_ g\) on the space \({\mathcal M}(G)\) of bounded measures on G. The corresponding metric \(d_ g\) (termed ``smoothing metric'' by the authors) metrizes the topology of weak (star) convergence in \({\mathcal M}^+(G)\) (Theorems 2.2 and 2.3). In the second part of the paper the authors investigate the behaviour of \(d_ g(P,P_ n)\) where P is a probability measure on the circle group or on the real line and where \(P_ n\) denotes the n th empirical measure for P. In a special case the limiting distribution of the sequence \((n^{1/2}d_ g(P,P_ n))_{n\geq 1}\) is determined making use of a CLT for empirical processes. Reviewer's remarks: 1. In Lemma 2.4 (and in the proof of Theorem 2.5) one has to require \(f\neq 0.\) 2. The proof of Theorem 2.2, (2)\(\to (3)\), is immediate taking into account Theorem 2.5; Lemma 2.6 is not needed and might even be misleading since \(\mu_ n\to^{w}\nu\) yields no information on the sequence \((\| \mu_ n-\nu \|)_{n\geq 1}\) (besides its boundedness). 3. The proof of Theorem 2.3, (3)\(\to (2)\), is not by contradiction (as claimed) but is direct. Moreover one has to argue first for the adjoint measure \({\tilde \mu}\) of \(\mu\) (cf. the proof of Theorem 2.5). 4. In Corollary 2.5 it should read: If the translates of g span \(C_ 0(G)\) then \(g\in {\mathcal L}^ 1(G).\) 5. In the last line of Section 2 it should read \(d_ n(\mu,\nu)\).