A note on the normal approximation error for randomly weighted self-normalized sums (Q485552)

Let \(\boldsymbol{X}=\{X_{n}\}_{n\geq1}\) and \(\boldsymbol{Y}=\{Y_{n}\} _{n\geq1}\) be two independent random sequences. The authors investigate the rate of convergence to the normal distribution of the randomly weighted self-normalized sums \[ \psi_{n}=\psi_{n}(\boldsymbol{X}, \boldsymbol{Y})= \sum_{i=1}^{n} X_{i}Y_{i}/V_{n},\quad V_{n}=\sqrt{Y_{1}^{2}+\dots+Y_{n}^{2}}. \] The random variables \(\psi_{n}\) appear in some important statistics: for example, the Student \(t\)-statistic \(T_{n}=\psi_{n}[(n-1)/(n-\psi_{n}^{2})] ^{1/2}\), where \(\psi_{n}=\psi_{n}(\boldsymbol{1},\boldsymbol{Y})\) and the empirical correlation coefficient \(\rho_{n}(\boldsymbol{X},\boldsymbol{Y}) =\psi_{n}(\boldsymbol{X},\boldsymbol{Y})/B_{n}\), where \(B_{n}=\sqrt{X_{1}^{2}+\dots+X_{n}^{2}}\).