The law of the iterated logarithm of random weighting approximation for mean error -- non. I.I.D. situation (Q1367385)

Let \((X_n)\) be a sequence of independent random variables with \(EX_n= \mu\), \(n\geq 1\). Define \[ n\overline X= \sum^n_{j=1} X_jT_n=\overline X-\mu,\quad F_n(x)= P(T_n/(\text{var } T_n)^{1/2}\leq x), \] \[ H_n= \sum^n_{j=1} (X_j-\overline X)\eta_j,\quad\overline H^2_n= \sum^n_{j=1} (X_j-\overline X)^2,\quad F^*_n(x)= P(H_n/\overline H_n\leq x), \] \(\eta_1,\eta_2,\eta_3,\dots\) are i.i.d. \(\Gamma(4,2)\) random variables. Under some conditions on the centralized moments of \(X_n\) the authors establish that \[ \sup_{x\in R} |F(x_n)- F^*(x_n)|= O((\log\log n)^{1/2}/n). \] The technique used is to expand \(F_n(x)\) as an Edgeworth type of series, to expand \(F^*_n(x)\) as a series involving the standard normal distribution function and sums of powers of deviations \((X_j-\overline X)/\overline H_n\), \(j=1,2,\dots,n\), and to apply the classical form of the law of the iterated logarithm to such sums.