Convergence of two functionals of asymptotically normal sums of independent random variables (Q1273373)

Let \(X_{n,k}\), \(k= 1,\dots, m_n\), be a sequence of arrays of independent inside each row random variables. Denote \(a_{n,k}(\tau)\) and \(\sigma^2_{n,k}(\tau)\) the expectation and the variance of \(X_{n,k}\), respectively, truncated at level \(\tau\) where \(\tau\) is a fixed positive number. It is supposed that the random variables are uniformly asymptotically negligible. The author investigates the asymptotic behavior of the following two functionals: \[ \sum^{m_n}_{k= 1}\sigma^2_{n, k}(\tau) \left(\sum^k_{i= 1} X_{n,i}\right)^2 \quad\text{and}\quad \sum^{m_n}_{k= 1} p(t_{n, k}) \sigma^2_{n,k}(\tau) \left(\sum^k_{i= 1} (X_{n,i}- a_{n,i})\right)^2, \] where \(p(x)\), \(x\in [0,1]\), is a continuous function with at most numerals zeros, \[ t_{n,k}= \sum^k_{i=1} \sigma^2_{n,i}(\tau)\Biggl/ \sum^{m_n}_{i= 1} \sigma^2_{n, i}(\tau), \qquad k=1,\dots, m_n. \] It is proved that the central limit theorem for the sums \(\sum^{m_n}_{k= 1} (X_{n, k}-a_{n, k})\) and some additional conditions involve the weak convergence of these functionals. Theorem 1 asserts that the validity of the central limit theorem of mentioned sums is necessary and sufficient for the weak convergence of the second functional. It should be noted that only the sufficiency part is proved.