Functional central limit theorems for approximate martingale arrays in a locally compact Abelian group (Q1099866)

This paper is a continuation of two recent papers of the author on the central limit theorem on locally compact Abelian groups [Math. Z. 192, 409-419 (1986; Zbl 0599.60010) and ibid. 195, 1-12 (1987; Zbl 0633.60013)]. The investigation of functional central limit theorems for arrays of real-valued random variables subject to martingale conditions [cf. \textit{P. Hall} and \textit{C. C. Heyde}, Martingale limit theory and its application (1980; Zbl 0462.60045)] has inspired the present work for related arrays of random variables with values in a locally compact Abelian group G. If \(\{S_{nj}:\) \(1\leq j\leq k_ n\), \(n\geq 1\}\) is a triangular G-valued array with differences \(X_{nj}=S_{nj}-S_{n,j-1}\), then the random elements \(t\to S_ n(t):=\sum^{[k_ nt]}_{j=1}X_{nj}\), \(0\leq t\leq 1\), take values in the Skorokhod space D(G). Under the approximate martingale condition (due to the author), the uniform infinitesimality condition and some further familiar conditions, the sequence \((S_ n)\) converges stably in law, respectively weakly, to a probability distribution \(\mu\) on D(G) whose finite-dimensional distributions are mixtures of Gaussian distributions; moreover \(\mu\) is supported by the subspace C(G) of continuous functions in D(G) (Theorems 1 and 2). Finally, if \((\Phi_{s,t})_{0\leq s\leq t\leq 1}\) is a continuous hemigroup of continuous nonnegative quadratic forms on the dual group of G, then there exists a probability measure \(\mu\) on C(G) whose finite- dimensional distributions are canonically determined by the hemigroup (Theorem 3).