Asymptotic expansions in the conditional central limit theorem (Q2640217)

The authors prove the following theorem: Let \(X_ n\), \(n\in N\), be a sequence of real valued random variables with \[ E(X_ 1)=0,\quad E(X^ 2_ 1)=1,\quad E(| X_ 1|^ r)<\infty,\quad r\geq 3;\quad S_ n=\sum^{n}_{i=1}X_ i,\quad S^*_ n=S_ n/\sqrt{n}, \] g be a bounded measurable function, \(\Phi\) (t) be the standard normal distribution function, and \(\phi\) (t) be its density. If some complementary conditions are fulfilled, then there exist polynomials \(Q_{i,g}(t)\) such that \[ \sup_{t\in R}| E(S^*_ n\leq t,g)- \Phi (t)E(g)-\phi (t)\sum^{j(r,\beta)}_{i=1}Q_{i,g}(t)n^{- i/2}| =O_ n(r,\beta), \] where \(j(r,\beta)=[r]-2\) or r-3, and the estimate \(O_ n(r,\beta)\) has different values for different r and \(\beta\), in general \(O_ n=O(n^{-(r-2)/2\pm \epsilon})\), \(\epsilon >0\).