A new way to obtain estimates in the invariance principle (Q4524745)

The paper deals with the strong approximation of a sequence of independent random variables \(X_i\) of distributions \(F_i\) by another sequence of independent random variables \(Y_i\) of distributions \(G_i\). In the manner of \textit{I. Berkes} and \textit{W. Philipp} [Ann. Probab. 7, 29-54 (1979; Zbl 0392.60024)], the author constructs the \(Y_i\) sequentially on a single probability space in such a manner that the probability that the pointwise distance \(|X_i-Y_i|\) is non-zero is controlled by the distance between the corresponding distributions. The innovation here is the introduction of a further sequence \(V_i\) so that \(X_i+V_i\) is approximated by \(Y_i+V_{i-1}\). This permits a tighter approximation, and the extra terms telescope together when considering the partial sums of the \(X\) and \(Y\) sequences. The author applies this result to cases such as those where the \(F_i\) and \(G_i\) are known only up to a fixed number of moments, or infinite-dimensional martingales. For the entire collection see [Zbl 0948.00040].