A note on exact convergence rates in some martingale central limit theorems (Q2565370): Difference between revisions

\textit{E. Bolthausen} [Ann. Probab. 10, 672-688 (1982; Zbl 0494.60020)] has obtained a bound of order \(n^{-1/2}\) on the rate of convergence in the central limit theorem for arrays of martingale differences having uniformly bounded conditional moments of order four. Because of this moment condition the theorem falls short to include the classical Berry-Esséen theorem for independent and identically distributed random variables, but Bolthausen conjectured that an appropriate sharpening of his result should be possible. The present paper shows that this is not the case by providing an example of an array of martingale difference sequences which has uniformly bounded conditional moments of order three (even times on appropriate unbounded \(h\)-function) whereas the rate of convergence to normality is slightly worse than \(n^{-1/2}\). This example is part of a careful study about how much Bolthausen's moment condition can be relaxed while maintaining the rate of order \(n^{-1/2}\). Moreover, exact rates of convergence are derived under conditions on conditional moments of an order between two and three.