Central limit theorems revisited (Q1976508)

The authors prove four central limit theorems (CLT). The first CLT is for a triangular array of random elements taking values in a real separable Hilbert space, where the components of the array are row-wise independent and have finite second moments, the covariance operator of the row sum satisfies mild convergence requirements, and the array satisfies the Lindeberg condition along the evaluation at an orthonormal basis. The second CLT is for a real-valued martingale difference array for which the row sum of conditional variances converges in probability to a constant \(\sigma^{2}\) and a conditional Lindeberg condition is satisfied. The third CLT is an extension of the second one to multi-dimensions, and the fourth CLT is an extension of the third one to the Hilbert space setting. The first CLT is an extension of Theorem~4.3.2 of \textit{V. Fabian} and \textit{J. Hannan} [``Introduction to probability and mathematical statistics'' (1985; Zbl 0558.62001)]. The second CLT is Theorem~VIII.1 of \textit{D. Pollard} [``Convergence of stochastic processes'' (1984; Zbl 0544.60045)] strengthened to include \(\sigma^{2}=0\). The technique, namely convergence in distribution via uniform approximation, is used to extend the results from finite to infinite dimensions. Extension to random elements taking values in a Banach space with a Schauder basis is indicated.